545 U.S. 967 (2005)

[972] Paul T. Cappuccio argued the cause for petitioners in No. 04-277. With him on the briefs were Howard J. Symons, Tara M. Corvo, Paul Glist, John D. Seiver, David E. Mills, Daniel L. Brenner, Neal M. Goldberg, Michael S. Schooler, Edward J. Weiss, and Henk Brands.

Deputy Solicitor General Hungar argued the cause for federal petitioners in No. 04-281. With him on the briefs were Acting Solicitor General Clement, Assistant Attorney [973] General Pate, Deputy Assistant Attorney General Delrahim, James A. Feldman, Catherine G. O'Sullivan, Nancy C. Garrison, John A. Rogovin, Austin C. Schlick, Daniel M. Armstrong, Jacob M. Lewis, and Nandan M. Joshi.

Thomas C. Goldstein argued the cause for respondents in both cases. With him on the brief were Amy Howe, John W. Butler, Earl W. Comstock, Alison B. Macdonald, Harvey L. Reiter, Matthew J. Verschelden, and Andrew Jay Schwartzman. William H. Sorrell, Attorney General of Vermont, David Borsykowsky, Assistant Attorney General, and Ellen S. LeVine filed a brief in both cases for respondents State of Vermont et al. Michael K. Kellogg, Sean A. Lev, and James G. Harralson filed a brief in both cases for respondents BellSouth et al. Andrew G. McBride, Eve Klindera Reed, William P. Barr, Michael E. Glover, Edward Shakin, and John P. Frantz filed a brief in both cases for respondents Verizon Telephone Companies et al. Mark D. Schneider, Marc A. Goldman, and Jeffrey A. Rackow filed a brief in both cases for respondent MCI, Inc.^[2]

JUSTICE THOMAS delivered the opinion of the Court.

Title II of the Communications Act of 1934, 48 Stat. 1064, as amended, 47 U. S. C. § 151 et seq., subjects all providers of "telecommunications servic[e]" to mandatory common-carrier regulation, § 153(44). In the order under review, the [974] Federal Communications Commission concluded that cable companies that sell broadband Internet service do not provide "telecommunications servic[e]" as the Communications Act defines that term, and hence are exempt from mandatory common-carrier regulation under Title II. We must decide whether that conclusion is a lawful construction of the Communications Act under Chevron U. S. A. Inc. v. Natural Resources Defense Council, Inc., 467 U. S. 837 (1984), and the Administrative Procedure Act, 5 U. S. C. § 551 et seq. We hold that it is.

I

The traditional means by which consumers in the United States access the network of interconnected computers that make up the Internet is through "dial-up" connections provided over local telephone facilities. See 345 F. 3d 1120, 1123-1124 (CA9 2003) (cases below); In re Inquiry Concerning High-Speed Access to the Internet Over Cable and Other Facilities, 17 FCC Rcd. 4798, 4802-4803, ¶ 9 (2002) (hereinafter Declaratory Ruling). Using these connections, consumers access the Internet by making calls with computer modems through the telephone wires owned by local phone companies. See Verizon Communications Inc. v. FCC, 535 U. S. 467, 489-490 (2002) (describing the physical structure of a local telephone exchange). Internet service providers (ISPs), in turn, link those calls to the Internet network, not only by providing a physical connection, but also by offering consumers the ability to translate raw Internet data into information they may both view on their personal computers and transmit to other computers connected to the Internet. See In re Federal-State Joint Board on Universal Service, 13 FCC Rcd. 11501, 11531, ¶ 63 (1998) (hereinafter Universal Service Report or Report); P. Huber, M. Kellogg, & J. Thorne, Federal Telecommunications Law 988 (2d ed. 1999) (hereinafter Huber); 345 F. 3d, at 1123-1124. Technological limitations of local telephone wires, however, retard the speed at which data from the Internet may be transmitted [975] through end users' dial-up connections. Dial-up connections are therefore known as "narrowband," or slower speed, connections.

"Broadband" Internet service, by contrast, transmits data at much higher speeds. There are two principal kinds of broadband Internet service: cable modem service and Digital Subscriber Line (DSL) service. Cable modem service transmits data between the Internet and users' computers via the network of television cable lines owned by cable companies. See id., at 1124. DSL service provides high-speed access using the local telephone wires owned by local telephone companies. See WorldCom, Inc. v. FCC, 246 F. 3d 690, 692 (CADC 2001) (describing DSL technology). Cable companies and telephone companies can either provide Internet access directly to consumers, thus acting as ISPs themselves, or can lease their transmission facilities to independent ISPs that then use the facilities to provide consumers with Internet access. Other ways of transmitting high-speed Internet data into homes, including terrestrial and satellite-based wireless networks, are also emerging. Declaratory Ruling 4802, ¶ 6.

II

At issue in these cases is the proper regulatory classification under the Communications Act of broadband cable Internet service. The Act, as amended by the Telecommunications Act of 1996, 110 Stat. 56, defines two categories of regulated entities relevant to these cases: telecommunications carriers and information-service providers. The Act regulates telecommunications carriers, but not information service providers, as common carriers. Telecommunications carriers, for example, must charge just and reasonable, nondiscriminatory rates to their customers, 47 U. S. C. §§ 201-209, design their systems so that other carriers can interconnect with their communications networks, § 251(a)(1), and contribute to the federal "universal service" fund, § 254(d). [976] These provisions are mandatory, but the Commission must forbear from applying them if it determines that the public interest requires it. §§ 160(a), (b). Information-service providers, by contrast, are not subject to mandatory common-carrier regulation under Title II, though the Commission has jurisdiction to impose additional regulatory obligations under its Title I ancillary jurisdiction to regulate interstate and foreign communications, see §§ 151-161.

These two statutory classifications originated in the late 1970's, as the Commission developed rules to regulate data-processing services offered over telephone wires. That regime, the "Computer II" rules, distinguished between "basic" service (like telephone service) and "enhanced" service (computer-processing service offered over telephone lines). In re Amendment of Section 64.702 of the Commission's Rules and Regulations (Second Computer Inquiry), 77 F. C. C. 2d 384, 417-423, ¶¶ 86-101 (1980) (hereinafter Computer II Order). The Computer II rules defined both basic and enhanced services by reference to how the consumer perceives the service being offered.

In particular, the Commission defined "basic service" as "a pure transmission capability over a communications path that is virtually transparent in terms of its interaction with customer supplied information." Id., at 420, ¶ 96. By "pure" or "transparent" transmission, the Commission meant a communications path that enabled the consumer to transmit an ordinary-language message to another point, with no computer processing or storage of the information, other than the processing or storage needed to convert the message into electronic form and then back into ordinary language for purposes of transmitting it over the network— such as via a telephone or a facsimile. Id., at 419-420, ¶¶ 94-95. Basic service was subject to common-carrier regulation. Id., at 428, ¶ 114.

"[E]nhanced service," however, was service in which "computer processing applications [were] used to act on the [977] content, code, protocol, and other aspects of the subscriber's information," such as voice and data storage services, id., at 420-421, ¶ 97, as well as "protocol conversion" (i. e., ability to communicate between networks that employ different data-transmission formats), id., at 421-422, ¶ 99. By contrast to basic service, the Commission decided not to subject providers of enhanced service, even enhanced service offered via transmission wires, to Title II common-carrier regulation. Id., at 428-432, ¶¶ 115-123. The Commission explained that it was unwise to subject enhanced service to common-carrier regulation given the "fast-moving, competitive market" in which they were offered. Id., at 434, ¶ 129.

The definitions of the terms "telecommunications service" and "information service" established by the 1996 Act are similar to the Computer II basic- and enhanced-service classifications. "Telecommunications service"—the analog to basic service—is "the offering of telecommunications for a fee directly to the public . . . regardless of the facilities used." 47 U. S. C. § 153(46). "Telecommunications" is "the transmission, between or among points specified by the user, of information of the user's choosing, without change in the form or content of the information as sent and received." § 153(43). "Telecommunications carrier[s]"—those subjected to mandatory Title II common-carrier regulation—are defined as "provider[s] of telecommunications services." § 153(44). And "information service"—the analog to enhanced service—is "the offering of a capability for generating, acquiring, storing, transforming, processing, retrieving, utilizing, or making available information via telecommunications...." § 153(20).

In September 2000, the Commission initiated a rulemaking proceeding to, among other things, apply these classifications to cable companies that offer broadband Internet service directly to consumers. In March 2002, that rulemaking culminated in the Declaratory Ruling under review in these cases. In the Declaratory Ruling, the Commission concluded [978] that broadband Internet service provided by cable companies is an "information service" but not a "telecommunications service" under the Act, and therefore not subject to mandatory Title II common-carrier regulation. In support of this conclusion, the Commission relied heavily on its Universal Service Report. See Declaratory Ruling 4821-4822, ¶¶ 36-37 (citing Universal Service Report). The Universal Service Report classified "non-facilities-based" ISPs— those that do not own the transmission facilities they use to connect the end user to the Internet—solely as information-service providers. See Universal Service Report 11533, ¶ 67. Unlike those ISPs, cable companies own the cable lines they use to provide Internet access. Nevertheless, in the Declaratory Ruling, the Commission found no basis in the statutory definitions for treating cable companies differently from non-facilities-based ISPs: Both offer "a single, integrated service that enables the subscriber to utilize Internet access service . . . and to realize the benefits of a comprehensive service offering." Declaratory Ruling 4823, ¶ 38. Because Internet access provides a capability for manipulating and storing information, the Commission concluded that it was an information service. Ibid.

The integrated nature of Internet access and the high-speed wire used to provide Internet access led the Commission to conclude that cable companies providing Internet access are not telecommunications providers. This conclusion, the Commission reasoned, followed from the logic of the Universal Service Report. The Report had concluded that, though Internet service "involves data transport elements" because "an Internet access provider must enable the movement of information between customers' own computers and distant computers with which those customers seek to interact," it also "offers end users information-service capabilities inextricably intertwined with data transport." Universal Service Report 11539-11540, ¶ 80. ISPs, therefore, were not "offering . . . telecommunications . . . directly to the public," [979] § 153(46), and so were not properly classified as telecommunications carriers, see id., at 11540, ¶ 81. In other words, the Commission reasoned that consumers use their cable modems not to transmit information "transparently," such as by using a telephone, but instead to obtain Internet access.

The Commission applied this same reasoning to cable companies offering broadband Internet access. Its logic was that, like non-facilities-based ISPs, cable companies do not "offe[r] telecommunications service to the end user, but rather . . . merely us[e] telecommunications to provide end users with cable modem service." Declaratory Ruling 4824, ¶ 41. Though the Commission declined to apply mandatory Title II common-carrier regulation to cable companies, it invited comment on whether under its Title I jurisdiction it should require cable companies to offer other ISPs access to their facilities on common-carrier terms. Id., at 4839, ¶ 72. Numerous parties petitioned for judicial review, challenging the Commission's conclusion that cable modem service was not telecommunications service. By judicial lottery, the Court of Appeals for the Ninth Circuit was selected as the venue for the challenge.

The Court of Appeals granted the petitions in part, vacated the Declaratory Ruling in part, and remanded to the Commission for further proceedings. In particular, the Court of Appeals vacated the ruling to the extent it concluded that cable modem service was not "telecommunications service" under the Communications Act. It held that the Commission could not permissibly construe the Communications Act to exempt cable companies providing Internet service from Title II regulation. See 345 F. 3d, at 1132. Rather than analyzing the permissibility of that construction under the deferential framework of Chevron, 467 U. S. 837, however, the Court of Appeals grounded its holding in the stare decisis effect of AT&T; Corp. v. Portland, 216 F. 3d 871 (CA9 2000). See 345 F. 3d, at 1128-1132. Portland held that cable modem service was a "telecommunications service," [980] though the court in that case was not reviewing an administrative proceeding and the Commission was not a party to the case. See 216 F. 3d, at 877-880. Nevertheless, Portland's holding, the Court of Appeals reasoned, overrode the contrary interpretation reached by the Commission in the Declaratory Ruling. See 345 F. 3d, at 1130-1131.

We granted certiorari to settle the important questions of federal law that these cases present. 543 U. S. 1018 (2004).

III

We first consider whether we should apply Chevron's framework to the Commission's interpretation of the term "telecommunications service." We conclude that we should. We also conclude that the Court of Appeals should have done the same, instead of following the contrary construction it adopted in Portland.

A

In Chevron, this Court held that ambiguities in statutes within an agency's jurisdiction to administer are delegations of authority to the agency to fill the statutory gap in reasonable fashion. Filling these gaps, the Court explained, involves difficult policy choices that agencies are better equipped to make than courts. 467 U. S., at 865-866. If a statute is ambiguous, and if the implementing agency's construction is reasonable, Chevron requires a federal court to accept the agency's construction of the statute, even if the agency's reading differs from what the court believes is the best statutory interpretation. Id., at 843-844, and n. 11.

The Chevron framework governs our review of the Commission's construction. Congress has delegated to the Commission the authority to "execute and enforce" the Communications Act, § 151, and to "prescribe such rules and regulations as may be necessary in the public interest to carry out the provisions" of the Act, § 201(b); AT&T; Corp. v. Iowa Utilities Bd., 525 U. S. 366, 377-378 (1999). These provisions give the Commission the authority to promulgate [981] binding legal rules; the Commission issued the order under review in the exercise of that authority; and no one questions that the order is within the Commission's jurisdiction. See Household Credit Services, Inc. v. Pfennig, 541 U. S. 232, 238-239 (2004); United States v. Mead Corp., 533 U. S. 218, 231-234 (2001); Christensen v. Harris County, 529 U. S. 576, 586-588 (2000). Hence, as we have in the past, we apply the Chevron framework to the Commission's interpretation of the Communications Act. See National Cable & Telecommunications Assn., Inc. v. Gulf Power Co., 534 U. S. 327, 333-339 (2002); Verizon, 535 U. S., at 501-502.

Some of the respondents dispute this conclusion, on the ground that the Commission's interpretation is inconsistent with its past practice. We reject this argument. Agency inconsistency is not a basis for declining to analyze the agency's interpretation under the Chevron framework. Un-explained inconsistency is, at most, a reason for holding an interpretation to be an arbitrary and capricious change from agency practice under the Administrative Procedure Act. See Motor Vehicle Mfrs. Assn. of United States, Inc. v. State Farm Mut. Automobile Ins. Co., 463 U. S. 29, 46-57 (1983). For if the agency adequately explains the reasons for a reversal of policy, "change is not invalidating, since the whole point of Chevron is to leave the discretion provided by the ambiguities of a statute with the implementing agency." Smiley v. Citibank (South Dakota), N. A., 517 U. S. 735, 742 (1996); see also Rust v. Sullivan, 500 U. S. 173, 186-187 (1991); Barnhart v. Walton, 535 U. S. 212, 226 (2002) (SCALIA, J., concurring in part and concurring in judgment). "An initial agency interpretation is not instantly carved in stone. On the contrary, the agency ... must consider varying interpretations and the wisdom of its policy on a continuing basis," Chevron, supra, at 863-864, for example, in response to changed factual circumstances, or a change in administrations, see State Farm, supra, at 59 (REHNQUIST, J., concurring in part and dissenting in part). That is no doubt why [982] in Chevron itself, this Court deferred to an agency interpretation that was a recent reversal of agency policy. See 467 U. S., at 857-858. We therefore have no difficulty concluding that Chevron applies.

B

The Court of Appeals declined to apply Chevron because it thought the Commission's interpretation of the Communications Act foreclosed by the conflicting construction of the Act it had adopted in Portland. See 345 F. 3d, at 1127-1132. It based that holding on the assumption that Portland's construction overrode the Commission's, regardless of whether Portland had held the statute to be unambiguous. 345 F. 3d, at 1131. That reasoning was incorrect.

A court's prior judicial construction of a statute trumps an agency construction otherwise entitled to Chevron deference only if the prior court decision holds that its construction follows from the unambiguous terms of the statute and thus leaves no room for agency discretion. This principle follows from Chevron itself. Chevron established a "presumption that Congress, when it left ambiguity in a statute meant for implementation by an agency, understood that the ambiguity would be resolved, first and foremost, by the agency, and desired the agency (rather than the courts) to possess whatever degree of discretion the ambiguity allows." Smiley, supra, at 740-741. Yet allowing a judicial precedent to foreclose an agency from interpreting an ambiguous statute, as the Court of Appeals assumed it could, would allow a court's interpretation to override an agency's. Chevron's premise is that it is for agencies, not courts, to fill statutory gaps. See 467 U. S., at 843-844, and n. 11. The better rule is to hold judicial interpretations contained in precedents to the same demanding Chevron step one standard that applies if the court is reviewing the agency's construction on a blank slate: Only a judicial precedent holding that the statute [983] unambiguously forecloses the agency's interpretation, and therefore contains no gap for the agency to fill, displaces a conflicting agency construction.

A contrary rule would produce anomalous results. It would mean that whether an agency's interpretation of an ambiguous statute is entitled to Chevron deference would turn on the order in which the interpretations issue: If the court's construction came first, its construction would prevail, whereas if the agency's came first, the agency's construction would command Chevron deference. Yet whether Congress has delegated to an agency the authority to interpret a statute does not depend on the order in which the judicial and administrative constructions occur. The Court of Appeals' rule, moreover, would "lead to the ossification of large portions of our statutory law," Mead, 533 U. S., at 247 (Scalia, J., dissenting), by precluding agencies from revising unwise judicial constructions of ambiguous statutes. Neither Chevron nor the doctrine of stare decisis requires these haphazard results.

The dissent answers that allowing an agency to override what a court believes to be the best interpretation of a statute makes "judicial decisions subject to reversal by executive officers." Post, at 1016 (opinion of SCALIA, J.). It does not. Since Chevron teaches that a court's opinion as to the best reading of an ambiguous statute an agency is charged with administering is not authoritative, the agency's decision to construe that statute differently from a court does not say that the court's holding was legally wrong. Instead, the agency may, consistent with the court's holding, choose a different construction, since the agency remains the authoritative interpreter (within the limits of reason) of such statutes. In all other respects, the court's prior ruling remains binding law (for example, as to agency interpretations to which Chevron is inapplicable). The precedent has not been "reversed" by the agency, any more than a federal court's interpretation of a State's law can be said to have been "reversed" by a [984] state court that adopts a conflicting (yet authoritative) interpretation of state law.

The Court of Appeals derived a contrary rule from a mistaken reading of this Court's decisions. It read Neal v. United States, 516 U. S. 284 (1996), to establish that a prior judicial construction of a statute categorically controls an agency's contrary construction. 345 F. 3d, at 1131-1132; see also post, at 1016, n. 11 (SCALIA, J., dissenting). Neal established no such proposition. Neal declined to defer to a construction adopted by the United States Sentencing Commission that conflicted with one the Court previously had adopted in Chapman v. United States, 500 U. S. 453 (1991). Neal, supra, at 290-295. Chapman, however, had held the relevant statute to be unambiguous. See 500 U. S., at 463 (declining to apply the rule of lenity given the statute's clear language). Thus, Neal established only that a precedent holding a statute to be unambiguous forecloses a contrary agency construction. That limited holding accorded with this Court's prior decisions, which had held that a court's interpretation of a statute trumps an agency's under the doctrine of stare decisis only if the prior court holding "determined a statute's clear meaning." Maislin Industries, U. S., Inc. v. Primary Steel, Inc., 497 U. S. 116, 131 (1990) (emphasis added); see also Lechmere, Inc. v. NLRB, 502 U. S. 527, 536-537 (1992). Those decisions allow a court's prior interpretation of a statute to override an agency's interpretation only if the relevant court decision held the statute unambiguous.

Against this background, the Court of Appeals erred in refusing to apply Chevron to the Commission's interpretation of the definition of "telecommunications service," 47 U. S. C. § 153(46). Its prior decision in Portland held only that the best reading of § 153(46) was that cable modem service was a "telecommunications service," not that it was the only permissible reading of the statute. See 216 F. 3d, at 877-880. Nothing in Portland held that the Communications [985] Act unambiguously required treating cable Internet providers as telecommunications carriers. Instead, the court noted that it was "not presented with a case involving potential deference to an administrative agency's statutory construction pursuant to the Chevron doctrine," id., at 876; and the court invoked no other rule of construction (such as the rule of lenity) requiring it to conclude that the statute was unambiguous to reach its judgment. Before a judicial construction of a statute, whether contained in a precedent or not, may trump an agency's, the court must hold that the statute unambiguously requires the court's construction. Portland did not do so.

As the dissent points out, it is not logically necessary for us to reach the question whether the Court of Appeals misapplied Chevron for us to decide whether the Commission acted lawfully. See post, at 1019-1020 (opinion of SCALIA, J.). Nevertheless, it is no "great mystery" why we are reaching the point here. Post, at 1019. There is genuine confusion in the lower courts over the interaction between the Chevron doctrine and stare decisis principles, as the petitioners informed us at the certiorari stage of this litigation. See Pet. for Cert. of Federal Communications Commission et al. in No. 04-281, pp. 19-23; Pet. for Cert. of National Cable & Telecomm. Assn. et al. in No. 04-277, pp. 22-29. The point has been briefed. See Brief for Federal Petitioners 38-44; Brief for Cable-Industry Petitioners 30-36. And not reaching the point could undermine the purpose of our grant of certiorari: to settle authoritatively whether the Commission's Declaratory Ruling is lawful. Were we to uphold the Declaratory Ruling without reaching the Chevron point, the Court of Appeals could once again strike down the Commission's rule based on its Portland decision. Portland (at least arguably) could compel the Court of Appeals once again to reverse the Commission despite our decision, since our conclusion that it is reasonable to read the Communications Act to classify cable modem service solely as an "information [986] service" leaves untouched Portland's holding that the Commission's interpretation is not the best reading of the statute. We have before decided similar questions that were not, strictly speaking, necessary to our disposition. See, e. g., Agostini v. Felton, 521 U. S. 203, 237 (1997) (requiring the Courts of Appeals to adhere to our directly controlling precedents, even those that rest on reasons rejected in other decisions); Roper v. Simmons, 543 U. S. 551, 628-629 (2005) (SCALIA, J., dissenting) (criticizing this Court for not reaching the question whether the Missouri Supreme Court erred by failing to follow directly controlling Supreme Court precedent, though that conclusion was not necessary to the Court's decision). It is prudent for us to do so once again today.

IV

We next address whether the Commission's construction of the definition of "telecommunications service," 47 U. S. C. § 153(46), is a permissible reading of the Communications Act under the Chevron framework. Chevron established a familiar two-step procedure for evaluating whether an agency's interpretation of a statute is lawful. At the first step, we ask whether the statute's plain terms "directly addres[s] the precise question at issue." 467 U. S., at 843. If the statute is ambiguous on the point, we defer at step two to the agency's interpretation so long as the construction is "a reasonable policy choice for the agency to make." Id., at 845. The Commission's interpretation is permissible at both steps.

A

We first set forth our understanding of the interpretation of the Communications Act that the Commission embraced. The issue before the Commission was whether cable companies providing cable modem service are providing a "telecommunications service" in addition to an "information service."

[987] The Commission first concluded that cable modem service is an "information service," a conclusion unchallenged here. The Act defines "information service" as "the offering of a capability for generating, acquiring, storing, transforming, processing, retrieving, utilizing, or making available information via telecommunications . . . ." § 153(20). Cable modem service is an information service, the Commission reasoned, because it provides consumers with a comprehensive capability for manipulating information using the Internet via high-speed telecommunications. That service enables users, for example, to browse the World Wide Web, to transfer files from file archives available on the Internet via the "File Transfer Protocol," and to access e-mail and Usenet newsgroups. Declaratory Ruling 4821, ¶ 37; Universal Service Report 11537, ¶ 76. Like other forms of Internet service, cable modem service also gives users access to the Domain Name System (DNS). DNS, among other things, matches the Web page addresses that end users type into their browsers (or "click" on) with the Internet Protocol (IP) addresses^[3] of the servers containing the Web pages the users wish to access. Declaratory Ruling 4821-4822, ¶ 37. All of these features, the Commission concluded, were part of the information service that cable companies provide consumers. Id., at 4821-4823, ¶¶ 36-38; see also Universal Service Report 11536-11539, ¶¶ 75-79.

At the same time, the Commission concluded that cable modem service was not "telecommunications service." "Telecommunications service" is "the offering of telecommunications for a fee directly to the public." 47 U. S. C. § 153(46). "Telecommunications," in turn, is defined as "the transmission, between or among points specified by the user, of information of the user's choosing, without change in the form or content of the information as sent and received." [988] § 153(43). The Commission conceded that, like all information-service providers, cable companies use "telecommunications" to provide consumers with Internet service; cable companies provide such service via the high-speed wire that transmits signals to and from an end user's computer. Declaratory Ruling 4823, ¶ 40. For the Commission, however, the question whether cable broadband Internet providers "offer" telecommunications involved more than whether telecommunications was one necessary component of cable modem service. Instead, whether that service also includes a telecommunications "offering" "turn[ed] on the nature of the functions the end user is offered," id., at 4822, ¶ 38 (emphasis added), for the statutory definition of "telecommunications service" does not "res[t] on the particular types of facilities used," id., at 4821, ¶ 35; see § 153(46) (definition of "telecommunications service" applies "regardless of the facilities used").

Seen from the consumer's point of view, the Commission concluded, cable modem service is not a telecommunications offering because the consumer uses the high-speed wire always in connection with the information-processing capabilities provided by Internet access, and because the transmission is a necessary component of Internet access: "As provided to the end user the telecommunications is part and parcel of cable modem service and is integral to its other capabilities." Declaratory Ruling 4823, ¶ 39. The wire is used, in other words, to access the World Wide Web, newsgroups, and so forth, rather than "transparently" to transmit and receive ordinary-language messages without computer processing or storage of the message. See supra, at 976 (noting the Computer II notion of "transparent" transmission). The integrated character of this offering led the Commission to conclude that cable modem service is not a "stand-alone," transparent offering of telecommunications. Declaratory Ruling 4823-4825, ¶¶ 41-43.

[989] B

This construction passes Chevron's first step. Respondents argue that it does not, on the ground that cable companies providing Internet service necessarily "offe[r]" the underlying telecommunications used to transmit that service. The word "offering" as used in § 153(46), however, does not unambiguously require that result. Instead, "offering" can reasonably be read to mean a "stand-alone" offering of telecommunications, i. e., an offered service that, from the user's perspective, transmits messages unadulterated by computer processing. That conclusion follows not only from the ordinary meaning of the word "offering," but also from the regulatory history of the Communications Act.

1

Cable companies in the broadband Internet service business "offe[r]" consumers an information service in the form of Internet access and they do so "via telecommunications," § 153(20), but it does not inexorably follow as a matter of ordinary language that they also "offe[r]" consumers the high-speed data transmission (telecommunications) that is an input used to provide this service, § 153(46). We have held that where a statute's plain terms admit of two or more reasonable ordinary usages, the Commission's choice of one of them is entitled to deference. See Verizon, 535 U. S., at 498 (deferring to the Commission's interpretation of the term "cost" by reference to an alternative linguistic usage defined by what "[a] merchant who is asked about `the cost of providing the goods'" might "reasonably" say); National Railroad Passenger Corporation v. Boston & Maine Corp., 503 U. S. 407, 418 (1992) (agency construction entitled to deference where there were "alternative dictionary definitions of the word" at issue). The term "offe[r]" as used in the definition of telecommunications service, § 153(46), is ambiguous in this way.

[990] It is common usage to describe what a company "offers" to a consumer as what the consumer perceives to be the integrated finished product, even to the exclusion of discrete components that compose the product, as the dissent concedes. See post, at 1006-1007 (opinion of Scalia, J.). One might well say that a car dealership "offers" cars, but does not "offer" the integrated major inputs that make purchasing the car valuable, such as the engine or the chassis. It would, in fact, be odd to describe a car dealership as "offering" consumers the car's components in addition to the car itself. Even if it is linguistically permissible to say that the car dealership "offers" engines when it offers cars, that shows, at most, that the term "offer," when applied to a commercial transaction, is ambiguous about whether it describes only the offered finished product, or the product's discrete components as well. It does not show that no other usage is permitted.

The question, then, is whether the transmission component of cable modem service is sufficiently integrated with the finished service to make it reasonable to describe the two as a single, integrated offering. See ibid. We think that they are sufficiently integrated, because "[a] consumer uses the high-speed wire always in connection with the information-processing capabilities provided by Internet access, and because the transmission is a necessary component of Internet access." Supra, at 988. In the telecommunications context, it is at least reasonable to describe companies as not "offering" to consumers each discrete input that is necessary to providing, and is always used in connection with, a finished service. We think it no misuse of language, for example, to say that cable companies providing Internet service do not "offer" consumers DNS, even though DNS is essential to providing Internet access. Declaratory Ruling 4810, n. 74, 4822-4823, ¶ 38. Likewise, a telephone company "offers" consumers a transparent transmission path that conveys an ordinary-language message, not necessarily the data-transmission [991] facilities that also "transmi[t] . . . information of the user's choosing," § 153(43), or other physical elements of the facilities used to provide telephone service, like the trunks and switches, or the copper in the wires. What cable companies providing cable modem service and telephone companies providing telephone service "offer" is Internet service and telephone service respectively—the finished services, though they do so using (or "via") the discrete components composing the end product, including data transmission. Such functionally integrated components need not be described as distinct "offerings."

In response, the dissent argues that the high-speed transmission component necessary to providing cable modem service is necessarily "offered" with Internet service because cable modem service is like the offering of pizza delivery service together with pizza, and the offering of puppies together with dog leashes. Post, at 1007-1008 (opinion of SCALIA, J.). The dissent's appeal to these analogies only underscores that the term "offer" is ambiguous in the way that we have described. The entire question is whether the products here are functionally integrated (like the components of a car) or functionally separate (like pets and leashes). That question turns not on the language of the Act, but on the factual particulars of how Internet technology works and how it is provided, questions Chevron leaves to the Commission to resolve in the first instance. As the Commission has candidly recognized, "the question may not always be straightforward whether, on the one hand, an entity is providing a single information service with communications and computing components, or, on the other hand, is providing two distinct services, one of which is a telecommunications service." Universal Service Report 11530, ¶ 60. Because the term "offer" can sometimes refer to a single, finished product and sometimes to the "individual components in a package being offered" (depending on whether the components "still possess sufficient identity to be described [992] as separate objects," post, at 1006), the statute fails unambiguously to classify the telecommunications component of cable modem service as a distinct offering. This leaves federal telecommunications policy in this technical and complex area to be set by the Commission, not by warring analogies.

We also do not share the dissent's certainty that cable modem service is so obviously like pizza delivery service and the combination of dog leashes and dogs that the Commission could not reasonably have thought otherwise. Post, at 1007-1008. For example, unlike the transmission component of Internet service, delivery service and dog leashes are not integral components of the finished products (pizzas and pet dogs). One can pick up a pizza rather than having it delivered, and one can own a dog without buying a leash. By contrast, the Commission reasonably concluded, a consumer cannot purchase Internet service without also purchasing a connection to the Internet and the transmission always occurs in connection with information processing. In any event, we doubt that a statute that, for example, subjected offerors of "delivery" service (such as Federal Express and United Parcel Service) to common-carrier regulation would unambiguously require pizza-delivery companies to offer their delivery services on a common-carrier basis.

2

The Commission's traditional distinction between basic and enhanced service, see supra, at 976-977, also supports the conclusion that the Communications Act is ambiguous about whether cable companies "offer" telecommunications with cable modem service. Congress passed the definitions in the Communications Act against the background of this regulatory history, and we may assume that the parallel terms "telecommunications service" and "information service" substantially incorporated their meaning, as the Commission has held. See, e. g., In re Federal-State Joint Board on Universal Service, 12 FCC Rcd. 8776, 9179-9180, ¶ 788 [993] (1997) (noting that the "definition of enhanced services is substantially similar to the definition of information services" and that "all services previously considered `enhanced services' are `information services'"); Commissioner v. Keystone Consol. Industries, Inc., 508 U. S. 152, 159 (1993) (noting presumption that Congress is aware of "settled judicial and administrative interpretation[s]" of terms when it enacts a statute). The regulatory history in at least two respects confirms that the term "telecommunications service" is ambiguous.

First, in the Computer II Order that established the terms "basic" and "enhanced" services, the Commission defined those terms functionally, based on how the consumer interacts with the provided information, just as the Commission did in the order below. See supra, at 976-977. As we have explained, Internet service is not "transparent in terms of its interaction with customer supplied information," Computer II Order 420, ¶ 96; the transmission occurs in connection with information processing. It was therefore consistent with the statute's terms for the Commission to assume that the parallel term "telecommunications service" in 47 U. S. C. § 153(46) likewise describes a "pure" or "transparent" communications path not necessarily separately present, from the end user's perspective, in an integrated information-service offering.

The Commission's application of the basic/enhanced-service distinction to non-facilities-based ISPs also supports this conclusion. The Commission has long held that "all those who provide some form of transmission services are not necessarily common carriers." Computer II Order 431, ¶ 122; see also id., at 435, ¶ 132 ("acknowledg[ing] the existence of a communications component" in enhanced-service offerings). For example, the Commission did not subject to common-carrier regulation those service providers that offered enhanced services over telecommunications facilities, but that did not themselves own the underlying facilities— so-called "non-facilities-based" providers. See Universal [994] Service Report 11530, ¶ 60. Examples of these services included database services in which a customer used telecommunications to access information, such as Dow Jones News and Lexis, as well as "value added networks," which lease wires from common carriers and provide transmission as well as protocol-processing service over those wires. See In re Amendment to Sections 64.702 of the Commission's Rules and Regulations (Third Computer Inquiry), 3 FCC Rcd. 1150, 1153, n. 23 (1988); supra, at 977 (explaining protocol conversion). These services "combin[ed] communications and computing components," yet the Commission held that they should "always be deemed enhanced" and therefore not subject to common-carrier regulation. Universal Service Report 11530, ¶ 60. Following this traditional distinction, the Commission in the Universal Service Report classified ISPs that leased rather than owned their transmission facilities as pure information-service providers. Id., at 11540, ¶ 81.

Respondents' statutory arguments conflict with this regulatory history. They claim that the Communications Act unambiguously classifies as telecommunications carriers all entities that use telecommunications inputs to provide information service. As respondent MCI concedes, this argument would subject to mandatory common-carrier regulation all information-service providers that use telecommunications as an input to provide information service to the public. Brief for Respondent MCI, Inc., 30. For example, it would subject to common-carrier regulation non-facilities-based ISPs that own no transmission facilities. See Universal Service Report 11532-11533, ¶ 66. Those ISPs provide consumers with transmission facilities used to connect to the Internet, see supra, at 974, and so, under respondents' argument, necessarily "offer" telecommunications to consumers. Respondents' position that all such entities are necessarily "offering telecommunications" therefore entails mandatory common-carrier regulation of entities that the Commission [995] never classified as "offerors" of basic transmission service, and therefore common carriers, under the Computer II regime.^[4] See Universal Service Report 11540, ¶ 81 (noting past Commission policy); Computer and Communications Industry Assn. v. FCC, 693 F. 2d 198, 209 (CADC 1982) (noting and upholding Commission's Computer II "finding that enhanced services . . . are not common carrier services within the scope of Title II"). We doubt that the parallel term "telecommunications service" unambiguously worked this abrupt shift in Commission policy.

Respondents' analogy between cable companies that provide cable modem service and facilities-based enhanced-service providers—that is, enhanced-service providers who own the transmission facilities used to provide those services—fares no better. Respondents stress that under the Computer II rules the Commission regulated such providers more heavily than non-facilities-based providers. The Commission required, for example, local telephone companies that provided enhanced services to offer their wires on a common-carrier basis to competing enhanced-service providers. See, e. g., In re Amendment of Sections 64.702 of the Commission's Rules and Regulations (Third Computer Inquiry), 104 F. C. C. 2d 958, 964, ¶ 4 (1986) (hereinafter Computer III Order). Respondents argue that the Communications Act unambiguously requires the same treatment for cable companies because cable companies also own the facilities they use to provide cable modem service (and therefore information service).

[996] We disagree. We think it improbable that the Communications Act unambiguously freezes in time the Computer II treatment of facilities-based information-service providers. The Act's definition of "telecommunications service" says nothing about imposing more stringent regulatory duties on facilities-based information-service providers. The definition hinges solely on whether the entity "offer[s] telecommunications for a fee directly to the public," 47 U. S. C. § 153(46), though the Act elsewhere subjects facilities-based carriers to stricter regulation, see § 251(c) (imposing various duties on facilities-based local telephone companies). In the Computer II rules, the Commission subjected facilities-based providers to common-carrier duties not because of the nature of the "offering" made by those carriers, but rather because of the concern that local telephone companies would abuse the monopoly power they possessed by virtue of the "bottleneck" local telephone facilities they owned. See Computer II Order 474-475, ¶¶ 229, 231; Computer III Order 968-969, ¶ 12; Verizon, 535 U. S., at 489-490 (describing the naturally monopolistic physical structure of a local telephone exchange). The differential treatment of facilities-based carriers was therefore a function not of the definitions of "enhanced-service" and "basic service," but instead of a choice by the Commission to regulate more stringently, in its discretion, certain entities that provided enhanced service. The Act's definitions, however, parallel the definitions of enhanced and basic service, not the facilities-based grounds on which that policy choice was based, and the Commission remains free to impose special regulatory duties on facilities-based ISPs under its Title I ancillary jurisdiction. In fact, it has invited comment on whether it can and should do so. See supra, at 979.

In sum, if the Act fails unambiguously to classify nonfacilities-based information-service providers that use telecommunications inputs to provide an information service as "offer[ors]" of "telecommunications," then it also fails unambiguously [997] to classify facilities-based information-service providers as telecommunications-service offerors; the relevant definitions do not distinguish facilities-based and nonfacilities-based carriers. That silence suggests, instead, that the Commission has the discretion to fill the consequent statutory gap.

C

We also conclude that the Commission's construction was "a reasonable policy choice for the [Commission] to make" at Chevron's second step. 467 U. S., at 845.

Respondents argue that the Commission's construction is unreasonable because it allows any communications provider to "evade" common-carrier regulation by the expedient of bundling information service with telecommunications. Respondents argue that under the Commission's construction a telephone company could, for example, offer an information service like voice mail together with telephone service, thereby avoiding common-carrier regulation of its telephone service.

We need not decide whether a construction that resulted in these consequences would be unreasonable because we do not believe that these results follow from the construction the Commission adopted. As we understand the Declaratory Ruling, the Commission did not say that any telecommunications service that is priced or bundled with an information service is automatically unregulated under Title II. The Commission said that a telecommunications input used to provide an information service that is not "separable from the data-processing capabilities of the service" and is instead "part and parcel of [the information service] and is integral to [the information service's] other capabilities" is not a telecommunications offering. Declaratory Ruling 4823, ¶ 39; see supra, at 988.

This construction does not leave all information-service offerings exempt from mandatory Title II regulation. "It is plain," for example, that a local telephone company "cannot [998] escape Title II regulation of its residential local exchange service simply by packaging that service with voice mail." Universal Service Report 11530, ¶ 60. That is because a telephone company that packages voice mail with telephone service offers a transparent transmission path—telephone service—that transmits information independent of the information-storage capabilities provided by voice mail. For instance, when a person makes a telephone call, his ability to convey and receive information using the call is only trivially affected by the additional voice-mail capability. Equally, were a telephone company to add a time-of-day announcement that played every time the user picked up his telephone, the "transparent" information transmitted in the ensuing call would be only trivially dependent on the information service the announcement provides. By contrast, the high-speed transmission used to provide cable modem service is a functionally integrated component of that service because it transmits data only in connection with the further processing of information and is necessary to provide Internet service. The Commission's construction therefore was more limited than respondents assume.

Respondents answer that cable modem service does, in fact, provide "transparent" transmission from the consumer's perspective, but this argument, too, is mistaken. Respondents characterize the "information-service" offering of Internet access as consisting only of access to a cable company's e-mail service, its Web page, and the ability it provides consumers to create a personal Web page. When a consumer goes beyond those offerings and accesses content provided by parties other than the cable company, respondents argue, the consumer uses "pure transmission" no less than a consumer who purchases phone service together with voice mail.

This argument, we believe, conflicts with the Commission's understanding of the nature of cable modem service, an understanding we find to be reasonable. When an end user [999] accesses a third-party's Web site, the Commission concluded, he is equally using the information service provided by the cable company that offers him Internet access as when he accesses the company's own Web site, its e-mail service, or his personal Web page. For example, as the Commission found below, part of the information service cable companies provide is access to DNS service. See supra, at 987. A user cannot reach a third-party's Web site without DNS, which (among other things) matches the Web site address the end user types into his browser (or "clicks" on with his mouse) with the IP address of the Web page's host server. See P. Albitz & C. Liu, DNS and BIND 10 (4th ed. 2001) (For an Internet user, "DNS is a must. . . . [N]early all of the Internet's network services use DNS. That includes the World Wide Web, electronic mail, remote terminal access, and file transfer"). It is at least reasonable to think of DNS as a "capability for . . . acquiring . . . retrieving, utilizing, or making available" Web site addresses and therefore part of the information service cable companies provide. 47 U. S. C. § 153(20).^[5] Similarly, the Internet service provided by cable companies facilitates access to third-party Web pages by offering consumers the ability to store, or "cache," popular content on local computer servers. See Declaratory Ruling 4810, ¶ 17, and n. 76. Cacheing obviates the need for the end user to download anew information from third-party [1000] Web sites each time the consumer attempts to access them, thereby increasing the speed of information retrieval. In other words, subscribers can reach third-party Web sites via "the World Wide Web, and browse their contents, [only] because their service provider offers the `capability for . . . acquiring, [storing] . . . retrieving [and] utilizing . . . information.'" Universal Service Report 11538, ¶ 76 (quoting 47 U. S. C. § 153(20)). "The service that Internet access providers offer to members of the public is Internet access," Universal Service Report 11539, ¶ 79, not a transparent ability (from the end user's perspective) to transmit information. We therefore conclude that the Commission's construction was reasonable.

V

Respondent MCI, Inc., urges that the Commission's treatment of cable modem service is inconsistent with its treatment of DSL service, see supra, at 975 (describing DSL service), and therefore is an arbitrary and capricious deviation from agency policy. See 5 U. S. C. § 706(2)(A). MCI points out that when local telephone companies began to offer Internet access through DSL technology in addition to telephone service, the Commission applied its Computer II facilities-based classification to them and required them to make the telephone lines used to transmit DSL service available to competing ISPs on nondiscriminatory, common-carrier terms. See supra, at 996 (describing Computer II facilities-based classification of enhanced-service providers); In re Deployment of Wireline Services Offering Advanced Telecommunications Capability, 13 FCC Rcd. 24011, 24030-24031, ¶¶ 36-37 (1998) (hereinafter Wireline Order) (classifying DSL service as a telecommunications service). MCI claims that the Commission's decision not to regulate cable companies similarly under Title II is inconsistent with its DSL policy.

We conclude, however, that the Commission provided a reasoned explanation for treating cable modem service differently [1001] from DSL service. As we have already noted, see supra, at 981-982, the Commission is free within the limits of reasoned interpretation to change course if it adequately justifies the change.^[6] It has done so here. The traditional reason for its Computer II common-carrier treatment of facilities-based carriers (including DSL carriers), as the Commission explained, was "that the telephone network [was] the primary, if not exclusive, means through which information service providers can gain access to their customers." Declaratory Ruling 4825, ¶ 44 (emphasis in original; internal quotation marks omitted). The Commission applied the same treatment to DSL service based on that history, rather than on an analysis of contemporaneous market conditions. See Wireline Order 24031, ¶ 37 (noting DSL carriers' "continuing obligation" to offer their transmission facilities to competing ISPs on nondiscriminatory terms).

The Commission in the order under review, by contrast, concluded that changed market conditions warrant different treatment of facilities-based cable companies providing Internet access. Unlike at the time of Computer II, substitute forms of Internet transmission exist today: "[R]esidential high-speed access to the Internet is evolving over multiple electronic platforms, including wireline, cable, terrestrial wireless and satellite." Declaratory Ruling 4802, ¶ 6; see also U. S. Telecom Assn. v. FCC, 290 F. 3d 415, 428 (CADC 2002) (noting Commission findings of "robust competition . . . in the broadband market"). The Commission concluded that "`broadband services should exist in a minimal regulatory environment that promotes investment and innovation in a competitive market.'" Declaratory Ruling 4802, ¶ 5. [1002] This, the Commission reasoned, warranted treating cable companies unlike the facilities-based enhanced-service providers of the past. Id., at 4825, ¶ 44. We find nothing arbitrary about the Commission's providing a fresh analysis of the problem as applied to the cable industry, which it has never subjected to these rules. This is adequate rational justification for the Commission's conclusions.

Respondents argue, in effect, that the Commission's justification for exempting cable modem service providers from common-carrier regulation applies with similar force to DSL providers. We need not address that argument. The Commission's decision appears to be a first step in an effort to reshape the way the Commission regulates information-service providers; that may be why it has tentatively concluded that DSL service provided by facilities-based telephone companies should also be classified solely as an information service. See In re Appropriate Framework for Broadband Access to the Internet over Wireline Facilities, 17 FCC Rcd. 3019, 3030, ¶ 20 (2002). The Commission need not immediately apply the policy reasoning in the Declaratory Ruling to all types of information-service providers. It apparently has decided to revisit its longstanding Computer II classification of facilities-based information-service providers incrementally. Any inconsistency between the order under review and the Commission's treatment of DSL service can be adequately addressed when the Commission fully reconsiders its treatment of DSL service and when it decides whether, pursuant to its ancillary Title I jurisdiction, to require cable companies to allow independent ISPs access to their facilities. See supra, at 979 and this page. We express no view on those matters. In particular, we express no view on how the Commission should, or lawfully may, classify DSL service.

* * *

The questions the Commission resolved in the order under review involve a "subject matter [that] is technical, complex, [1003] and dynamic." Gulf Power, 534 U. S., at 339. The Commission is in a far better position to address these questions than we are. Nothing in the Communications Act or the Administrative Procedure Act makes unlawful the Commission's use of its expert policy judgment to resolve these difficult questions. The judgment of the Court of Appeals is reversed, and the cases are remanded for further proceedings consistent with this opinion.

It is so ordered.

JUSTICE STEVENS, concurring.

While I join the Court's opinion in full, I add this caveat concerning Part III-B, which correctly explains why a court of appeals' interpretation of an ambiguous provision in a regulatory statute does not foreclose a contrary reading by the agency. That explanation would not necessarily be applicable to a decision by this Court that would presumably remove any pre-existing ambiguity.

JUSTICE BREYER, concurring.

I join the Court's opinion because I believe that the Federal Communications Commission's decision falls within the scope of its statutorily delegated authority—though perhaps just barely. I write separately because I believe it important to point out that JUSTICE SCALIA, in my view, has wrongly characterized the Court's opinion in United States v. Mead Corp., 533 U. S. 218 (2001). He states that the Court held in Mead that "some unspecified degree of formal process" before the agency "was required" for courts to accord the agency's decision deference under Chevron U. S. A. Inc. v. Natural Resources Defense Council, Inc., 467 U. S. 837 (1984). Post, at 1015 (dissenting opinion); see also ibid. (formal process is "at least the only safe harbor").

JUSTICE SCALIA has correctly characterized the way in which he, in dissent, characterized the Court's Mead opinion. 533 U. S., at 245-246. But the Court said the opposite. An [1004] agency action qualifies for Chevron deference when Congress has explicitly or implicitly delegated to the agency the authority to "fill" a statutory "gap," including an interpretive gap created through an ambiguity in the language of a statute's provisions. Chevron, supra, at 843-844; Mead, supra, at 226-227. The Court said in Mead that such delegation "may be shown in a variety of ways, as by an agency's power to engage in adjudication or notice-and-comment rulemaking, or by some other indication of a comparable congressional intent." 533 U. S., at 227 (emphasis added). The Court explicitly stated that the absence of notice-and-comment rulemaking did "not decide the case," for the Court has "sometimes found reasons for Chevron deference even when no such administrative formality was required and none was afforded." Id., at 231. And the Court repeated that it "has recognized a variety of indicators that Congress would expect Chevron deference." Id., at 237 (emphasis added).

It is not surprising that the Court would hold that the existence of a formal rulemaking proceeding is neither a necessary nor a sufficient condition for according Chevron deference to an agency's interpretation of a statute. It is not a necessary condition because an agency might arrive at an authoritative interpretation of a congressional enactment in other ways, including ways that JUSTICE SCALIA mentions. See, e. g., Mead, supra, at 231. It is not a sufficient condition because Congress may have intended not to leave the matter of a particular interpretation up to the agency, irrespective of the procedure the agency uses to arrive at that interpretation, say, where an unusually basic legal question is at issue. Cf. General Dynamics Land Systems, Inc. v. Cline, 540 U. S. 581, 600 (2004) (rejecting agency's answer to question whether age discrimination law forbids discrimination against the relatively young).

Thus, while I believe JUSTICE SCALIA is right in emphasizing that Chevron deference may be appropriate in the absence [1005] of formal agency proceedings, Mead should not give him cause for concern.

JUSTICE SCALIA, with whom JUSTICE SOUTER and JUSTICE GINSBURG join as to Part I, dissenting.

The Federal Communications Commission (FCC or Commission) has once again attempted to concoct "a whole new regime of regulation (or of free-market competition)" under the guise of statutory construction. MCI Telecommunications Corp. v. American Telephone & Telegraph Co., 512 U. S. 218, 234 (1994). Actually, in these cases, it might be more accurate to say the Commission has attempted to establish a whole new regime of non-regulation, which will make for more or less free-market competition, depending upon whose experts are believed. The important fact, however, is that the Commission has chosen to achieve this through an implausible reading of the statute, and has thus exceeded the authority given it by Congress.

I

The first sentence of the FCC ruling under review reads as follows: "Cable modem service provides high-speed access to the Internet, as well as many applications or functions that can be used with that access, over cable system facilities." In re Inquiry Concerning High-Speed Access to the Internet Over Cable and Other Facilities, 17 FCC Rcd. 4798, 4799, ¶ 1 (2002) (hereinafter Declaratory Ruling) (emphasis added; footnote omitted). Does this mean that cable companies "offer" high-speed access to the Internet? Surprisingly not, if the Commission and the Court are to be believed.

It happens that cable-modem service is popular precisely because of the high-speed access it provides, and that, once connected with the Internet, cable-modem subscribers often use Internet applications and functions from providers other than the cable company. Nevertheless, for purposes of classifying [1006] what the cable company does, the Commission (with the Court's approval) puts all the emphasis on the rest of the package (the additional "applications or functions"). It does so by claiming that the cable company does not "offe[r]" its customers high-speed Internet access because it offers that access only in conjunction with particular applications and functions, rather than "separate[ly]," as a "stand-alone offering." Id., at 4802, ¶ 7, 4823, ¶ 40.

The focus on the term "offer" appropriately derives from the statutory definitions at issue in these cases. Under the Telecommunications Act of 1996, 110 Stat. 59, "`information service'" involves the capacity to generate, store, interact with, or otherwise manipulate "information via telecommunications." 47 U. S. C. § 153(20). In turn, "`telecommunications'" is defined as "the transmission, between or among points specified by the user, of information of the user's choosing, without change in the form or content of the information as sent and received." § 153(43). Finally, "`telecommunications service'" is defined as "the offering of telecommunications for a fee directly to the public . . . regardless of the facilities used." § 153(46). The question here is whether cable-modem-service providers "offe[r] . . . telecommunications for a fee directly to the public." If so, they are subject to Title II regulation as common carriers, like their chief competitors who provide Internet access through other technologies.

The Court concludes that the word "offer" is ambiguous in the sense that it has "`alternative dictionary definitions'" that might be relevant. Ante, at 989 (quoting National Railroad Passenger Corporation v. Boston & Maine Corp., 503 U. S. 407, 418 (1992)). It seems to me, however, that the analytic problem pertains not really to the meaning of "offer," but to the identity of what is offered. The relevant question is whether the individual components in a package being offered still possess sufficient identity to be described as separate objects of the offer, or whether they have been [1007] so changed by their combination with the other components that it is no longer reasonable to describe them in that way.

Thus, I agree (to adapt the Court's example, ante, at 990) that it would be odd to say that a car dealer is in the business of selling steel or carpets because the cars he sells include both steel frames and carpeting. Nor does the water company sell hydrogen, nor the pet store water (though dogs and cats are largely water at the molecular level). But what is sometimes true is not, as the Court seems to assume, always true. There are instances in which it is ridiculous to deny that one part of a joint offering is being offered merely because it is not offered on a "`stand-alone'" basis, ante, at 989.

If, for example, I call up a pizzeria and ask whether they offer delivery, both common sense and common "usage," ante, at 990, would prevent them from answering: "No, we do not offer delivery—but if you order a pizza from us, we'll bake it for you and then bring it to your house." The logical response to this would be something on the order of, "so, you do offer delivery." But our pizza-man may continue to deny the obvious and explain, paraphrasing the FCC and the Court: "No, even though we bring the pizza to your house, we are not actually `offering' you delivery, because the delivery that we provide to our end users is `part and parcel' of our pizzeria-pizza-at-home service and is `integral to its other capabilities.'" Cf. Declaratory Ruling 4823, ¶ 39; ante, at 988, 997-998.^[7] Any reasonable customer would conclude at that point that his interlocutor was either crazy or following some too-clever-by-half legal advice.

In short, for the inputs of a finished service to qualify as the objects of an "offer" (as that term is reasonably understood), it is perhaps a sufficient, but surely not a necessary, condition that the seller offer separately "each discrete input [1008] that is necessary to providing . . . a finished service," ante, at 990. The pet store may have a policy of selling puppies only with leashes, but any customer will say that it does offer puppies—because a leashed puppy is still a puppy, even though it is not offered on a "stand-alone" basis.

Despite the Court's mighty labors to prove otherwise, ante, at 989-1000, the telecommunications component of cable-modem service retains such ample independent identity that it must be regarded as being on offer—especially when seen from the perspective of the consumer or the end user, which the Court purports to find determinative, ante, at 990, 993, 998, 1000. The Commission's ruling began by noting that cable-modem service provides both "high-speed access to the Internet" and other "applications and functions," Declaratory Ruling 4799, ¶ 1, because that is exactly how any reasonable consumer would perceive it: as consisting of two separate things.

The consumer's view of the matter is best assessed by asking what other products cable-modem service substitutes for in the marketplace. Broadband Internet service provided by cable companies is one of the three most common forms of Internet service, the other two being dial-up access and broadband Digital Subscriber Line (DSL) service. Ante, at 974-975. In each of the other two, the physical transmission pathway to the Internet is sold—indeed, is legally required to be sold—separately from the Internet functionality. With dial-up access, the physical pathway comes from the telephone company, and the Internet service provider (ISP) provides the functionality.

"In the case of Internet access, the end user utilizes two different and distinct services. One is the transmission pathway, a telecommunications service that the end user purchases from the telephone company. The second is the Internet access service, which is an enhanced service provided by an ISP. . . . Th[e] functions [provided by the ISP] are separate from the transmission pathway [1009] over which that data travels. The pathway is a regulated telecommunications service; the enhanced service offered over it is not." FCC, Office of Plans and Policy, J. Oxman, The FCC and the Unregulation of the Internet, p. 13 (Working Paper No. 31, July 1999), available at http://www.fcc.gov/Bureaus/OPP/working_papers/oppwp31.pdf (as visited June 24, 2005, and available in Clerk of Court's case file).^[8]

As the Court acknowledges, ante, at 1000, DSL service has been similar to dial-up service in the respect that the physical connection to the Internet must be offered separately from Internet functionality.^[9] Thus, customers shopping for dial-up or DSL service will not be able to use the Internet unless they get both someone to provide them with a physical connection and someone to provide them with applications and functions such as e-mail and Web access. It is therefore inevitable that customers will regard the competing cable-modem service as giving them both computing functionality and the physical pipe by which that functionality comes to their computer—both the pizza and the delivery service that nondelivery pizzerias require to be purchased from the cab company.^[10]

[1010] Since the delivery service provided by cable (the broadband connection between the customer's computer and the cable company's computer-processing facilities) is downstream from the computer-processing facilities, there is no question that it merely serves as a conduit for the information services that have already been "assembled" by the cable company in its capacity as ISP. This is relevant because of the statutory distinction between an "information service" and "telecommunications." The former involves the capability of getting, processing, and manipulating information. § 153(20). The latter, by contrast, involves no "change in the form or content of the information as sent and received." § 153(43). When cable-company-assembled information enters the cable for delivery to the subscriber, the information service is already complete. The information has been (as the statute requires) generated, acquired, stored, transformed, processed, retrieved, utilized, or made available. All that remains is for the information in its final, unaltered form, to be delivered (via telecommunications) to the subscriber.

This reveals the insubstantiality of the fear invoked by both the Commission and the Court: the fear of what will happen to ISPs that do not provide the physical pathway to Internet access, yet still use telecommunications to acquire the pieces necessary to assemble the information that they pass back to their customers. According to this reductio, ante, at 993-995, if cable-modem-service providers are deemed to provide "telecommunications service," then so must all ISPs because they all "use" telecommunications in providing Internet functionality (by connecting to other [1011] parts of the Internet, including Internet backbone providers, for example). In terms of the pizzeria analogy, this is equivalent to saying that, if the pizzeria "offers" delivery, all restaurants "offer" delivery, because the ingredients of the food they serve their customers have come from other places; no matter how their customers get the food (whether by eating it at the restaurant, or by coming to pick it up themselves), they still consume a product for which delivery was a necessary "input." This is nonsense. Concluding that delivery of the finished pizza constitutes an "offer" of delivery does not require the conclusion that the serving of prepared food includes an "offer" of delivery. And that analogy does not even do the point justice, since "`telecommunications service'" is defined as "the offering of telecommunications for a fee directly to the public." § 153(46) (emphasis added). The ISPs' use of telecommunications in their processing of information is not offered directly to the public.

The "regulatory history" on which the Court depends so much, ante, at 992-997, provides another reason why common-carrier regulation of all ISPs is not a worry. Under its Computer Inquiry rules, which foreshadowed the definitions of "information" and "telecommunications" services, ante, at 976-977, the Commission forbore from regulating as common carriers "value-added networks"—non-facilities-based providers who leased basic services from common carriers and bundled them with enhanced services; it said that they, unlike facilities-based providers, would be deemed to provide only enhanced services, ante, at 993-994.^[11] That [1012] same result can be achieved today under the Commission's statutory authority to forbear from imposing most Title II regulations. § 160. In fact, the statutory criteria for forbearance—which include what is "just and reasonable," "necessary for the protection of consumers," and "consistent with the public interest," §§ 160(a)(1), (2), (3)—correspond well with the kinds of policy reasons the Commission has invoked to justify its peculiar construction of "telecommunications service" to exclude cable-modem service.

The Court also puts great stock in its conclusion that cable-modem subscribers cannot avoid using information services provided by the cable company in its ISP capacity, even when they only click-through to other ISPs. Ante, at 998-1000. For, even if a cable-modem subscriber uses e-mail from another ISP, designates some page not provided by the cable company as his home page, and takes advantage of none of the other standard applications and functions provided by the cable company, he will still be using the cable company's Domain Name System (DNS) server and, when he goes to popular Web pages, perhaps versions of them that are stored in the cable company's cache. This argument suffers from at least two problems. First, in the context of telephone services, the Court recognizes a de minimis exception to contamination of a telecommunications service by an information service. Ante, at 997-998. A similar exception would seem to apply to the functions in question here. DNS, in particular, is scarcely more than routing information, [1013] which is expressly excluded from the definition of "information service." § 153(20).^[12] Second, it is apparently possible to sell a telecommunications service separately from, although in conjunction with, ISP-like services; that is precisely what happens in the DSL context, and the Commission does not contest that it could be done in the context of cable. The only impediment appears to be the Commission's failure to require from cable companies the unbundling that it required of facilities-based providers under its Computer Inquiry.

Finally, I must note that, notwithstanding the Commission's self-congratulatory paean to its deregulatory largesse, e. g., Brief for Federal Petitioners 29-32, it concluded the Declaratory Ruling by asking, as the Court paraphrases, "whether under its Title I jurisdiction [the Commission] should require cable companies to offer other ISPs access to their facilities on common-carrier terms." Ante, at 979; see also Reply Brief for Federal Petitioners 9; Tr. of Oral Arg. 17. In other words, what the Commission hath given, the Commission may well take away—unless it doesn't. This is a wonderful illustration of how an experienced agency can (with some assistance from credulous courts) turn statutory constraints into bureaucratic discretions. The main source of the Commission's regulatory authority over common carriers is Title II, but the Commission has rendered that inapplicable in this instance by concluding that the definition of "telecommunications service" is ambiguous and does not (in [1014] its current view) apply to cable-modem service. It contemplates, however, altering that (unnecessary) outcome, not by changing the law (i. e., its construction of the Title II definitions), but by reserving the right to change the facts. Under its undefined and sparingly used "ancillary" powers, the Commission might conclude that it can order cable companies to "unbundle" the telecommunications component of cable-modem service.^[13] And presto, Title II will then apply to them, because they will finally be "offering" telecommunications service! Of course, the Commission will still have the statutory power to forbear from regulating them under § 160 (which it has already tentatively concluded it would do, Declaratory Ruling 4847-4848, ¶¶ 94-95). Such Möbius-strip reasoning mocks the principle that the statute constrains the agency in any meaningful way.

After all is said and done, after all the regulatory cant has been translated, and the smoke of agency expertise blown away, it remains perfectly clear that someone who sells cable-modem service is "offering" telecommunications. For that simple reason set forth in the statute, I would affirm the Court of Appeals.

II

In Part III-B of its opinion, the Court continues the administrative-law improvisation project it began four years ago in United States v. Mead Corp., 533 U. S. 218 (2001). To the extent it set forth a comprehensible rule,^[14] Mead drastically [1015] limited the categories of agency action that would qualify for deference under Chevron U. S. A. Inc. v. Natural Resources Defense Council, Inc., 467 U. S. 837 (1984). For example, the position taken by an agency before the Supreme Court, with full approval of the agency head, would not qualify. Rather, some unspecified degree of formal process was required—or was at least the only safe harbor. See Mead, supra, at 245-246 (SCALIA, J., dissenting).^[15]

This meant that many more issues appropriate for agency determination would reach the courts without benefit of an agency position entitled to Chevron deference, requiring the courts to rule on these issues de novo.^[16] As I pointed out in [1016] dissent, this in turn meant (under the law as it was understood until today)^[17] that many statutory ambiguities that might be resolved in varying fashions by successive agency administrations would be resolved finally, conclusively, and forever, by federal judges—producing an "ossification of large portions of our statutory law," 533 U. S., at 247. The Court today moves to solve this problem of its own creation by inventing yet another breathtaking novelty: judicial decisions subject to reversal by executive officers.

Imagine the following sequence of events: FCC action is challenged as ultra vires under the governing statute; the litigation reaches all the way to the Supreme Court of the United States. The Solicitor General sets forth the FCC's official position (approved by the Commission) regarding interpretation of the statute. Applying Mead, however, the Court denies the agency position Chevron deference, finds that the best interpretation of the statute contradicts the agency's position, and holds the challenged agency action unlawful. The agency promptly conducts a rulemaking, and [1017] adopts a rule that comports with its earlier position—in effect disagreeing with the Supreme Court concerning the best interpretation of the statute. According to today's opinion, the agency is thereupon free to take the action that the Supreme Court found unlawful.

This is not only bizarre. It is probably unconstitutional. As we held in Chicago & Southern Air Lines, Inc. v. Waterman S. S. Corp., 333 U. S. 103 (1948), Article III courts do not sit to render decisions that can be reversed or ignored by executive officers. In that case, the Court of Appeals had determined it had jurisdiction to review an order of the Civil Aeronautics Board awarding an overseas air route. By statute such orders were subject to Presidential approval and the order in question had in fact been approved by the President. Id., at 110-111. In order to avoid any conflict with the President's foreign-affairs powers, the Court of Appeals concluded that it would review the board's action "as a regulatory agent of Congress," and the results of that review would remain subject to approval or disapproval by the President. Id., at 112-113. As I noted in my Mead dissent, 533 U. S., at 248, the Court bristled at the suggestion: "Judgments within the powers vested in courts by the Judiciary Article of the Constitution may not lawfully be revised, overturned or refused faith and credit by another Department of Government." Waterman, supra, at 113. That is what today's decision effectively allows. Even when the agency itself is party to the case in which the Court construes a statute, the agency will be able to disregard that construction and seek Chevron deference for its contrary construction the next time around.^[18]

[1018] Of course, like Mead itself, today's novelty in belated remediation of Mead creates many uncertainties to bedevil the lower courts. A court's interpretation is conclusive, the Court says, only if it holds that interpretation to be "the only permissible reading of the statute," and not if it merely holds it to be "the best reading." Ante, at 984. Does this mean that in future statutory-construction cases involving agency-administered statutes courts must specify (presumably in dictum) which of the two they are holding? And what of the many cases decided in the past, before this dictum's requirement was established? Apparently, silence on the point means that the court's decision is subject to agency reversal: "Before a judicial construction of a statute, whether contained in a precedent or not, may trump an agency's, the court must hold that the statute unambiguously requires the court's construction."^[19] Ante, at 985. (I have not made, and as far as I know the Court has not made, any calculation of how many hundreds of past statutory decisions are now agency-reversible because of failure to include an "unambiguous" finding. I suspect the number is very large.) How much extra work will it entail for each court confronted with an agency-administered statute to determine whether it has reached, not only the right ("best") result, but "the only permissible" result? Is the standard for "unambiguous" under the Court's new agency-reversal rule the same as the standard for "unambiguous" under step one of Chevron? (If so, [1019] of course, every case that reaches step two of Chevron will be agency-reversible.) Does the "unambiguous" dictum produce stare decisis effect even when a court is affirming, rather than reversing, agency action—so that in the future the agency must adhere to that affirmed interpretation? If so, does the victorious agency have the right to appeal a Court of Appeals judgment in its favor, on the ground that the text in question is in fact not (as the Court of Appeals held) unambiguous, so the agency should be able to change its view in the future?

It is indeed a wonderful new world that the Court creates, one full of promise for administrative-law professors in need of tenure articles and, of course, for litigators.^[20] I would adhere to what has been the rule in the past: When a court interprets a statute without Chevron deference to agency views, its interpretation (whether or not asserted to rest upon an unambiguous text) is the law. I might add that it is a great mystery why any of this is relevant here. Whatever the stare decisis effect of AT&T; Corp. v. Portland, 216 F. 3d 871 (CA9 2000), in the Ninth Circuit, it surely does not govern this Court's decision. And—despite the Court's peculiar, self-abnegating suggestion to the contrary, ante, at 985-986—the Ninth Circuit would already be obliged to [1020] abandon Portland's holding in the face of this Court's decision that the Commission's construction of "telecommunications service" is entitled to deference and is reasonable. It is a sadness that the Court should go so far out of its way to make bad law.

I respectfully dissent.

[1] Together with No. 04-281, Federal Communications Commission et al. v. Brand X Internet Services et al., also on certiorari to the same court.

[2] Briefs of amici curiae urging reversal in both cases were filed for the Telecommunications Industry Association by Colleen L. Boothby and Andrew M. Brown; and for the Washington Legal Foundation by Daniel J. Popeo and David Price.

Briefs of amici curiae urging affirmance in both cases were filed for the State of New Jersey, Board of Public Utilities, by Peter C. Harvey, Attorney General of New Jersey, Andrea M. Silkowitz, Assistant Attorney General, and Kenneth J. Sheehan, Deputy Attorney General; for AARP et al. by Stacy Canan and Michael Schuster; for the American Civil Liberties Union et al. by Steven R. Shapiro, Christopher A. Hansen, Jennifer Stisa Granick, and Marjorie Heins; and for the National Association of Regulatory Utility Commissioners by James Bradford Ramsay.

[3] IP addresses identify computers on the Internet, enabling data packets transmitted from other computers to reach them. See Universal Service Report 11531, ¶ 62; Huber 985.

[4] The dissent attempts to escape this consequence of respondents' position by way of an elaborate analogy between ISPs and pizzerias. Post, at 1011 (opinion of SCALIA, J.). This analogy is flawed. A pizzeria "delivers" nothing, but ISPs plainly provide transmission service directly to the public in connection with Internet service. For example, with dial-up service, ISPs process the electronic signal that travels over local telephone wires, and transmit it to the Internet. See supra, at 974-975; Huber 988. The dissent therefore cannot deny that its position logically would require applying presumptively mandatory Title II regulation to all ISPs.

[5] The dissent claims that access to DNS does not count as use of the information-processing capabilities of Internet service because DNS is "scarcely more than routing information, which is expressly excluded from the definition of `information service.'" Post, at 1012-1013, and n. 6 (opinion of SCALIA, J.). But the definition of information service does not exclude "routing information." Instead, it excludes "any use of any such capability for the management, control, or operation of a telecommunications system or the management of a telecommunications service." 47 U. S. C. § 153(20). The dissent's argument therefore begs the question because it assumes that Internet service is a "telecommunications system" or "service" that DNS manages (a point on which, contrary to the dissent's assertion, post, at 1013, n. 6, we need take no view for purposes of this response).

[6] Respondents vigorously argue that the Commission's purported inconsistent treatment is a reason for holding the Commission's construction impermissible under Chevron U. S. A. Inc. v. Natural Resources Defense Council, Inc., 467 U. S. 837 (1984). Any inconsistency bears on whether the Commission has given a reasoned explanation for its current position, not on whether its interpretation is consistent with the statute.

[7] The myth that the pizzeria does not offer delivery becomes even more difficult to maintain when the pizzeria advertises quick delivery as one of its advantages over competitors. That, of course, is the case with cable broadband.

[8] See also In re Federal-State Joint Board on Universal Service, 13 FCC Rcd. 11501, 11571-11572, ¶ 145 (1998) (end users "obtain telecommunications service from local exchange carriers, and then use information services provided by their Internet service provider and [Web site operators] in order to access [the Web]").

[9] In the DSL context, the physical connection is generally resold to the consumer by an ISP that has taken advantage of the telephone company's offer. The consumer knows very well, however, that the physical connection is a necessary component for Internet access which, just as in the dial-up context, is not provided by the ISP.

[10] The Court contends that this analogy is inapposite because one need not have a pizza delivered, ante, at 992, whereas one must purchase the cable connection in order to use cable's ISP functions. But the ISP functions provided by the cable company can be used without cable delivery— by accessing them from an Internet connection other than cable. The merger of the physical connection and Internet functions in cable's offerings has nothing to do with the "`inextricably intertwined,'" ante, at 978, nature of the two (like a car and its carpet), but is an artificial product of the cable company's marketing decision not to offer the two separately, so that the Commission could (by the Declaratory Ruling under review here) exempt it from common-carrier status.

[11] The Commission says forbearance cannot explain why value-added networks were not regulated as basic-service providers because it was not given the power to forbear until 1996. Reply Brief for Federal Petitioners 3-4, n. 1. It is true that when the Commission ruled on value-added networks, the statute did not explicitly provide for forbearance—any more than it provided for the categories of basic and enhanced services that the Computer Inquiry rules established, and through which the forbearance was applied. The D. C. Circuit, however, had long since recognized the Commission's discretionary power to "forbear from Title II regulation." Computer and Communications Industry Assn. v. FCC, 693 F. 2d 198, 212 (1982).

The Commission also says its Computer Inquiry rules should not apply to cable because they were developed in the context of telephone lines. Brief for Federal Petitioners 35-36; see also ante, at 996. But to the extent that the statute imported the Computer Inquiry approach, there is no basis for applying it differently to cable than to telephone lines, since the definition of "telecommunications service" applies "regardless of the facilities used." 47 U. S. C. § 153(46).

[12] The Court says that invoking this explicit exception from the definition of information services, which applies only to the "management, control, or operation of a telecommunications system or the management of a telecommunications service," § 153(20), begs the question whether cable-modem service includes a telecommunications service, ante, at 999, n. 3. I think not, and cite the exception only to demonstrate that the incidental functions do not prevent cable from including a telecommunications service if it otherwise qualifies. It is rather the Court that begs the question, saying that the exception cannot apply because cable is not a telecommunications service.

[13] Under the Commission's assumption that cable-modem-service providers are not providing "telecommunications services," there is reason to doubt whether it can use its Title I powers to impose common-carrier-like requirements, since § 153(44) specifically provides that a "telecommunications carrier shall be treated as a common carrier under this chapter only to the extent that it is engaged in providing telecommunications services" (emphasis added), and "this chapter" includes Titles I and II.

[14] For a description of the confusion Mead has produced, see Vermeule, Mead in the Trenches, 71 Geo. Wash. L. Rev. 347, 361 (2003) (concluding that "the Court has inadvertently sent the lower courts stumbling into a no-man's land"); Bressman, How Mead Has Muddled Judicial Review of Agency Action, 58 Vand. L. Rev. 1443, 1475 (2005) ("Mead has muddled judicial review of agency action").

[15] JUSTICE BREYER attempts to clarify Mead by repeating its formulations that the Court has "sometimes found reasons" to give Chevron deference in a (still-unspecified) "variety of ways" or because of a (still-unspecified) "variety of indicators," ante, at 1004 (concurring opinion) (internal quotation marks and emphasis omitted). He also notes that deference is sometimes inappropriate for reasons unrelated to the agency's process. Surprising those who thought the Court's decision not to defer to the agency in General Dynamics Land Systems, Inc. v. Cline, 540 U. S. 581 (2004), depended on its conclusion that there was "no serious question . . . about purely textual ambiguity" in the statute, id., at 600, JUSTICE BREYER seemingly attributes that decision to a still-underdeveloped exception to Chevron deference—one for "unusually basic legal question[s]," ante, at 1004. The Court today (thankfully) does not follow this approach: It bases its decision on what it sees as statutory ambiguity, ante, at 996-997, without asking whether the classification of cable-modem service is an "unusually basic legal question."

[16] It is true that, even under the broad basis for deference that I propose (viz., any agency position that plainly has the approval of the agency head, see United States v. Mead Corp., 533 U. S. 218, 256-257 (2001) (SCALIA, J., dissenting)), some interpretive matters will be decided de novo, without deference to agency views. This would be a rare occurrence, however, at the Supreme Court level—at least with respect to matters of any significance to the agency. Seeking to achieve 100% agency control of ambiguous provisions through the complicated method the Court proposes is not worth the incremental benefit.

[17] The Court's unanimous holding in Neal v. United States, 516 U. S. 284 (1996), plainly rejected the notion that any form of deference could cause the Court to revisit a prior statutory-construction holding: "Once we have determined a statute's meaning, we adhere to our ruling under the doctrine of stare decisis, and we assess an agency's later interpretation of the statute against that settled law." Id., at 295. The Court attempts to reinterpret this plain language by dissecting the cases Neal cited, noting that they referred to previous determinations of "`a statute's clear meaning.'" Lechmere, Inc. v. NLRB, 502 U. S. 527, 537 (1992) (quoting Maislin Industries, U. S., Inc. v. Primary Steel, Inc., 497 U. S. 116, 131 (1990)). But those cases reveal that today's focus on the term "clear" is revisionist. The oldest case in the chain using that word, Maislin Industries, did not rely on a prior decision that held the statute to be clear, but on a run-of-the-mill statutory interpretation contained in a 1908 decision. Id., at 130-131. When Maislin Industries referred to the Court's prior determination of "a statute's clear meaning," it was referring to the fact that the prior decision had made the statute clear, and was not conducting a retrospective inquiry into whether the prior decision had declared the statute itself to be clear on its own terms.

[18] The Court contends that no reversal of judicial holdings is involved, because "a court's opinion as to the best reading of an ambiguous statute . . . is not authoritative," ante, at 983. That fails to appreciate the difference between a de novo construction of a statute and a decision whether to defer to an agency's position, which does not even "purport to give the statute a judicial interpretation." Mead, supra, at 248 (SCALIA, J., dissenting). Once a court has decided upon its de novo construction of the statute, there no longer is a "different construction" that is "consistent with the court's holding," ante, at 983, and available for adoption by the agency.

[19] Suggestive of the same chaotic undermining of all prior judicial decisions that do not explicitly renounce ambiguity is the Court's explanation of why agency departure from a prior judicial decision does not amount to overruling: "[T]he agency may, consistent with the court's holding, choose a different construction, since the agency remains the authoritative interpreter (within the limits of reason) of [ambiguous] statutes [it is charged with administering]." Ibid.

[20] Further deossification may already be on the way, as the Court has hinted that an agency construction unworthy of Chevron deference may be able to trump one of our statutory-construction holdings. In Edelman v. Lynchburg College, 535 U. S. 106, 114 (2002), the Court found "no need to resolve any question of deference" because the Equal Employment Opportunity Commission's rule was "the position we would adopt even if . . . we were interpreting the statute from scratch." It nevertheless refused to say whether the agency's position was "the only one permissible." Id., at 114, n. 8 (internal quotation marks omitted). JUSTICE O'CONNOR appropriately "doubt[ed] that it is possible to reserve" the question whether a regulation is entitled to Chevron deference "while simultaneously maintaining . . . that the agency is free to change its interpretation" in the future. 535 U. S., at 122 (opinion concurring in judgment). In response, the Court cryptically said only that "not all deference is deference under Chevron." Id., at 114, n. 8.

VERIZON, Appellant v. FEDERAL COMMUNICATIONS COMMISSION, Appellee. Independent Telephone & Telecommunications Alliance, et al., Intervenors.

Nos. 11-1355, 11-1356.

United States Court of Appeals, District of Columbia Circuit.

Argued Sept. 9, 2013.

Decided Jan. 14, 2014.

Helgi C. Walker argued the cause for appellant/petitioner Verizon. With her on the briefs were Eve Klindera Reed, William S. Consovoy, Brett A. Shumate, Walter E. Dellinger, Anton Metlitsky, Samir C. Jain, Carl W. Northrup, Michael Lazarus, Andrew Morentz, Michael E. Glover, William H. Johnson, Stephen B. Kinnaird, and Mark A. Stachiw. John T. Scott III and Edward Shakin entered appearances.

Stephen B. Kinnaird, Carl W. Northrup, Michael Lazarus, Andrew Morentz, and Mark A. Stachiw were on the briefs for appellants/petitioners MetroPCS Communications, Inc., et al.

John P. Elwood, Sam Kazman, Randolph May, and Ilya Shapiro were on the brief for amici curiae The Competitive Enterprise -Institute, et al. in support of appellant/petitioner.

Russell P. Hanser, Bryan N. Tramont, and Quentin Riegel were on the brief for amicus curiae National Association of Manufacturers in support of appellant/petitioner.

Kenneth T. Cuccinelli, II, Attorney General, Office of the Attorney General for the Commonwealth of Virginia, E. Duncan Getchell, Jr., Solicitor General, and Wesley G. Russell, Jr., Deputy Attorney General, were on the brief for amici curiae The Commonwealth of Virginia, et al. in support of appellant/petitioner.

Sean A. Lev, General Counsel, Federal Communications Commission, argued the cause for appellee/respondent. With him on the briefs were Catherine G. O’Sullivan and Nickolai G. Levin, Attorneys, U.S. Department of Justice, Peter Karanjia, Deputy General Counsel, Federal Communications Commission, Jacob M. Lewis, Associate General Counsel, and Joel Marcus and Matthew J. Dunne, Counsel. Robert J. Wiggers, Attorney, U.S. Department of Justice, R. Craig Lawrence, Assistant U.S. Attorney, and Richard K. Welch, Deputy Associate General Counsel, Federal Communications Commission, entered appearances.

Pantelis Michalopoulos argued the cause for intervenors. With him on the brief were Stephanie A. Roy, Andrew W. Guhr, Henry Goldberg, David C. Bergmann, Kurt Matthew Rogers, and Brendan Daniel Kasper. Markham C. Erickson, Jeffrey J. Binder, Harold J. Feld and James B. Ramsay entered appearances.

Sean H. Donahue and David T. Goldberg were on the brief for amici curiae Reed Hundt, et al. in support of appel-lee/respondent.

E. Joshua Rosenkranz, Gabriel M. Ramsey, Thomas J. Gray, and Christina Von der Ahe were on the brief for amici curiae Venture Capital Investors in support of appellee/respondent.

Andrew Jay Sehwartzman was on the brief for amicus curiae Tim Wu in support of appellee/respondent.

John Blevins was on the brief for amici curiae Internet Engineers and Technologists in support of appellee/respondent.

Kevin S. Bankston and Emma J. Llansó were on the brief for amici curiae The Center for Democracy and Technology, et al. in support of appellee/respondent.

Before: ROGERS and TATEL, Circuit Judges, and SILBERMAN, Senior Circuit Judge.

Opinion for the Court filed by Circuit Judge TATEL.

Opinion concurring in part and dissenting in part filed by Senior Circuit Judge SILBERMAN.

TATEL, Circuit Judge.

For the second time in four years, we are confronted with a Federal Communications Commission effort to compel broadband providers to treat all Internet traffic the same regardless of source — or to require, as it is popularly known, “net neutrality.” In Comcast Corp. v. FCC, 600 F.3d 642 (D.C.Cir.2010), we held that the Commission had failed to cite any statutory authority that would justify its order compelling a broadband provider to adhere to open network management practices. After Comcast, the Commission issued the order challenged here — In re Preserving the Open Internet, 25 F.C.C.R. 17905 (2010) (“the Open Internet Order”) — which imposes disclosure, anti-blocking, and anti-discrimination requirements on broadband providers. As we explain in this opinion, the Commission has established that section 706 of the Telecommunications Act of 1996 vests it with affirmative authority to enact measures encouraging the deployment of broadband infrastructure. The Commission, we further hold, has reasonably interpreted section 706 to empower it to promulgate rules governing broadband providers’ treatment of Internet traffic, and its justification for the specific rules at issue here — that they will preserve and facilitate the “virtuous circle” of innovation that has driven the explosive growth of the Internet — is reasonable and supported by substantial evidence. That said, even though the Commission has general authority to regulate in this arena, it may not impose requirements that contravene express statutory mandates. Given that the Commission has chosen to classify broadband providers in a manner that exempts them from treatment as common carriers, the Communications Act expressly prohibits the Commission from nonetheless regulating them as such. Because the Commission has failed to establish that the anti-discrimination and anti-blocking rules do not impose per se common carrier obligations, we vacate those portions of the Open Internet Order.

I.

Understanding this case requires an understanding of the Internet, the Internet marketplace, and the history of the Commission’s regulation of that marketplace.

Four major participants in the Internet marketplace are relevant to the issues before us: backbone networks, broadband providers, edge providers, and end users. Backbone networks are interconnected, long-haul fiber-optic links and high-speed routers capable of transmitting vast amounts of data. See In re Verizon Communications Inc. and MCI, Inc. Applications for Approval of Transfer of Control, 20 F.C.C.R. 18433, 18493 ¶ 110 (2005). Internet users generally connect to these networks — and, ultimately, to one another — through local access providers like petitioner Verizon, who operate the “last-mile” transmission lines. See Open Internet Order, 25 F.C.C.R. at 17908, 17915 ¶¶ 7, 20. In the Internet’s early days-, most users connected to the Internet through dial-up connections over local telephone lines. See In re Inquiry Concerning High-Speed Access to the Internet Over Cable and. Other Facilities, 17 F.C.C.R. 4798, 4802-03 ¶ 9 (2002) (“Cable Broadband Order”). Today, access is generally furnished through “broadband,” i.e., high-speed communications technologies, such as cable modem service. See In re Inquiry Concerning the Deployment of Advanced Telecommunications Capability to All Americans in a Reasonable and Timely Fashion, 25 F.C.C.R. 9556, 9557, 9558-59 ¶¶ 1, 4 (2010) (“Sixth Broadband Deployment Report ”); 47 U.S.C. § 1302(d)(1). Edge providers are those who, like Amazon or Google, provide content, services, and applications over the Internet, while end users are those who consume edge providers’ content, services, and applications. See Open Internet Order, 25 F.C.C.R. at 17910 ¶ 13. To pull the whole picture together with a slightly oversimplified example: when an edge provider such as YouTube transmits some sort of content — say, a video of a cat — to an end user, that content is broken down into packets of information, which are carried by the edge provider’s local access provider to the backbone network, which transmits these packets to the end user’s local access provider, which, in turn, transmits the information to the end user, who then views and hopefully enjoys the cat.

These categories of entities are not necessarily mutually exclusive. For example, end users may often act as edge providers by creating and sharing content that is consumed by other end users, for instance by posting photos on Facebook. Similarly, broadband providers may offer content, applications, and services that compete with those furnished by edge providers. See Open Internet Order, 25 F.C.C.R. at 17915 ¶ 20.

Proponents of net neutrality — or, to use the Commission’s preferred term, “Internet openness” — worry about the relationship between broadband providers and edge providers. They fear that broadband providers might prevent their end-user subscribers from accessing certain edge providers altogether, or might degrade the quality of their end-user subscribers’ access to certain edge providers, either as a means of favoring their own competing content or services or to enable them to collect fees from certain edge providers. Thus, for example, a broadband provider like Comcast might limit its end-user subscribers’ ability to access the New York Times website if it wanted to spike traffic to its own news website, or it might degrade the quality of the connection to a search website like Bing if a ■ competitor like Google paid for prioritized access.

Since the advent of the Internet, the Commission has confronted the questions of whether and how it should regulate this communications network, which, generally speaking, falls comfortably within the Commission’s jurisdiction over “all interstate and foreign communications by wire or radio.” 47 U.S.C. § 152(a). One of the Commission’s early efforts occurred in 1980, when it adopted what is known as the Computer II regime. The Computer II rules drew a line between “basic” services, which were subject to regulation under Title II of the Communications Act of 1934 as common carrier services, see 47 U.S.C. §§ 201 et seq., and “enhanced” services, which were not. See In re Amendment of Section 61.702 of the Commission’s Rules and Regulations, 77 F.C.C.2d 384, 387 ¶¶ 5-7 (1980) (“Second Computer Inquiry ”). What distinguished “enhanced” services from “basic” services was the extent to which they involved the processing of information rather than simply its transmission. Id. at 420-21 ¶¶ 96-97. For example, the Commission characterized telephone service as a “basic” service, see id. at 419 ¶ 94, because it involved a “pure” transmission that was “virtually transparent in terms of its interaction with customer supplied information,” id. at 420 ¶ 96. Services that involved “computer processing applications ... used to act on the content, code, protocol, and other aspects of the subscriber’s information” — a definition that encompassed the services needed to connect an end user to the Internet — constituted enhanced services. Id. at 420 ¶ 97.

By virtue of their designation as common carriers, providers of basic services were subject to the duties that apply to such entities, including that they “furnish ... communication service upon reasonable request,” 47 U.S.C. § 201(a), engage in no “unjust or unreasonable discrimination in charges, practices, classifications, regulations, facilities, or services,” id. § 202(a), and charge “just and reasonable” rates, id. § 201(b). Although the Commission applied no such restrictions to purveyors of enhanced services, it imposed limitations on certain entities, like AT & T, which owned the transmission facilities over which enhanced services would be provided. Second Computer Inquiry, 77 F.C.C.2d at 473-74 ¶¶ 228-29. These restrictions included, most significantly, requirements that such entities offer enhanced services only through a completely separate corporate entity and that they offer their transmissions facilities to other enhanced service providers on a common carrier basis. Id.

For more than twenty years, the Commission applied some form of the Computer II regime to Internet services offered over telephone lines, then the predominant way in which most end users connected to the Internet. See, e.g., In re Appropriate Framework for Broadband Access to the Internet Over Wireline Facilities, 17 F.C.C.R. 3019, 3037-40 ¶¶ 36-42 (2002). Telephone companies that provided the actual wireline facilities over which information was transmitted were limited in the manner in which they could provide the enhanced services necessary to permit end users to access the Internet. Id. at 3040 ¶ 42. They were also required to permit third-party Internet Service Providers (ISPs), such as America Online, to access their wireline transmission facilities on a common carrier basis. Id.

It was against this background that Congress passed the Telecommunications Act of 1996, Pub.L. No. 104-104, 110 Stat. 56. Tracking the Computer II distinction between basic and enhanced services, the Act defines two categories of entities: telecommunications carriers, which provide the equivalent of basic services, and information-service providers, which provide the equivalent of enhanced services. 47 U.S.C. § 153(24), (50), (51), (53); see National Cable & Telecommunications Ass’n v. Brand X Internet Services, 545 U.S. 967, 976-77, 125 S.Ct. 2688, 162 L.Ed.2d 820 (2005). The Act subjects telecommunications carriers, but not information-service providers, to Title II common carrier regulation. 47 U.S.C. § 153(53); Brand X, 545 U.S. at 975-76, 125 S.Ct. 2688.

Pursuant to the Act, and paralleling its prior practice under the Computer II regime, the Commission then classified Digital Subscriber Line (DSL) services— broadband Internet service furnished over telephone lines — as “telecommunications services.” See In re Deployment of Wire-line Services Offering Advanced Telecommunications Capability, 13 F.C.C.R. 24012, 24014, 24029-30 ¶¶3, 35-36 (1998) (“Advanced Services Order”). DSL services, the Commission concluded, involved pure transmission technologies, and so were subject to Title II regulation. Id. at 24030-31 ¶ 35. A DSL provider could exempt its Internet access services, but not its transmission facilities themselves, from Title II common carrier restrictions only by operating them through a separate affiliate (i.e., a quasi-independent ISP). Id. at 24018 ¶ 13.

Four years later, however, the Commission took a different approach when determining how to regulate broadband service provided by cable companies. Instead of viewing cable broadband providers’ transmission and processing of information as distinct services, the Commission determined that cable broadband providers— even those that own and operate the underlying last-mile transmission facilities— provide a “single, integrated information service.” Cable Broadband Order, 17 F.C.C.R. at 4824 ¶ 41. Because cable broadband providers were thus not telecommunications' carriers at all, they were entirely exempt from Title II regulation. Id. at 4802 ¶ 7.

In National Cable & Telecommunications Ass’n v. Brand X Internet Services, 545 U.S. 967, 125 S.Ct. 2688, 162 L.Ed.2d 820 (2005), the Supreme Court upheld the Commission’s classification of cable broadband providers. The Court concluded that the Commission’s ruling represented a reasonable interpretation of the 1996 Telecommunications Act’s ambiguous provision defining telecommunications service, see id. at 991-92, 125 S.Ct. 2688, and that the Commission’s determination was entitled to deference notwithstanding its apparent inconsistency with the agency’s prior interpretation of that statute, see id. at 981, 1000-01,125 S.Ct. 2688.

Following Brand X, the Commission classified other types of broadband providers, such as DSL and wireless, which includes those offering broadband Internet service for cellular telephones, as information service providers exempt from Title II’s common carrier requirements. See In re Appropriate Framework for Broadband Access to the Internet Over Wireline Facilities, 20 F.C.C.R. 14853, 14862 ¶12 (2005) (“2005 Wireline Broadband Order”); In re Appropriate Regulatory Treatment for Broadband Access to the Internet Over Wireless Networks, 22 F.C.C.R. 5901, 5901-02 ¶ 1 (2007) (“Wireless Broadband Order”); In re United Power Line Council’s Petition for Declaratory Ruling Regarding the Classification of Broadband over Power Line Internet Access Service as an Information Service, 21 F.C.C.R. 13281, 13281 ¶ 1 (2006). Despite calls to revisit these classification orders, see, e.g., Open Internet Order, 25 F.C.C.R. at 18046 (concurring statement of Commissioner Copps), the Commission has yet to overrule them.

But even as the Commission exempted broadband providers from Title II common carrier obligations, it left open the possibility that it would nonetheless regulate these entities. In the Cable Broadband Order, for example, the Commission sought comment on whether and to what extent it should utilize the powers granted it under Title I of the Communications Act to impose restrictions on cable broadband providers. Cable Broadband Order, 17 F.C.C.R. at 4842 ¶ 77. Subsequently, in conjunction with the 2005 Wireline Broadband Order, the Commission issued a Policy Statement in which it signaled its intention to “preserve and promote the open and interconnected nature of the public Internet.” In re Appropriate Framework for Broadband Access to the Internet Over Wireline Facilities, 20 F.C.C.R. 14986, 14988 ¶4 (2005). The Commission announced that should it “see evidence that providers of telecommunications for Internet access or IP-enabled services are violating these principles,” it would “not hesitate to take action to address that conduct.” 2005 Wireline Broadband Order, 20 F.C.C.R. at 14904 ¶ 96.

The Commission did just that when, two years later, several subscribers to Com-cast’s cable broadband service complained that the company had interfered with their use of certain peer-to-peer networking applications. See In re Formal Complaint of Free Press and Public Knowledge Against Comcast Corp. for Secretly Degrading Peer-to-Peer Applications, 23 F.C.C.R. 13028 (2008) (“Comcast Order ”). Finding that Comcast’s impairment of these applications had “contravene[d] ... federal policy,” id. at 13052 ¶43, the Commission ordered the company to adhere to a new approach for managing bandwidth demand and to disclose the details of that approach, id. at 13059-60 ¶ 54. The Commission justified its order as an exercise of what courts term its “ancillary jurisdiction,” see id. at 13034-41 ¶¶ 14-22, a power that flows from the broad language of Communications Act section 4(i). See 47 U.S.C. § 154(i) (“The Commission may perform any and all acts, make such rules and regulations, and issue such orders, not inconsistent with this chapter, as may be necessary in the execution of its functions.”); see generally American Library Ass’n v. FCC, 406 F.3d 689, 700-03 (D.C.Cir.2005). We have held that the Commission may exercise such ancillary jurisdiction where two conditions are met: “(1) the Commission’s general jurisdictional grant under Title I covers the regulated subject and (2) the regulations are reasonably ancillary to the Commission’s effective performance of its statutorily mandated responsibilities.” American Library Ass’n, 406 F.3d at 691-92.

In Comcast, we vacated the Commission’s order, holding that the agency failed to demonstrate that it possessed authority to regulate broadband providers’ network management practices. 600 F.3d at 644. Specifically, we held that the Commission had identified no grant of statutory authority to which the Comcast Order was reasonably ancillary. Id. at 661. The Commission had principally invoked statutory provisions that, though setting forth congressional policy, delegated no actual regulatory authority. Id. at 651-58. These provisions, we concluded, were insufficient because permitting the agency to ground its exercise of ancillary jurisdiction in policy statements alone would contravene the “ ‘axiomatic’ principle that ‘administrative agencies may [act] only pursuant to authority delegated to them by Congress.’ ” Id. at 654 (alteration in original) (quoting American Library Ass’n, 406 F.3d at 691). We went on to reject the Commission’s invocation of a handful of other statutory provisions that, although they could “arguably be read to delegate regulatory authority,” id. at 658, provided no support for the precise order at issue, id. at 658-61.

While the Comcast matter was pending, the Commission sought comment on a set of proposed rules that, with some modifications, eventually became the rules at issue here. See In re Preserving the Open Internet, 24 F.C.C.R. 13064 (2009). In support, it relied on the same theory of ancillary jurisdiction it had asserted in the Comcast Order. See id. at 13099 ¶¶ 83-85. But after our decision in Comcast undermined that theory, the Commission sought comment on whether and to what extent it should reclassify broadband Internet services as telecommunications services. See In re Framework for Broadband Internet Service, 25 F.C.C.R. 7866, 7867 ¶2 (2010). Ultimately, however, rather than reclassifying broadband, the Commission adopted the Open Internet Order that-Verizon challenges here. See 25 F.C.C.R. 17905.

The Open Internet Order establishes two sets of “prophylactic rules” designed to “incorporate longstanding openness principles that are generally in line with current practices.” 25 F.C.C.R. at 17907 ¶4. One set of rules applies to “fixed” broadband providers — i.e., those furnishing residential broadband service and, more generally, Internet access to end users “primarily at fixed end points using stationary equipment.” Id. at 17934 ¶ 49. The other set of requirements applies to “mobile” broadband providers — i.e., those “serv[ing] end users primarily using mobile stations,” such as smart phones. Id.

The Order first imposes a transparency requirement on both fixed and mobile broadband providers. Id. at 17938 ¶ 56. They must “publicly discldse accurate information regarding the network management practices, performance, and commercial terms of [their] broadband Internet access services.” Id. at 17937 ¶ 54 (fixed providers); see also id. at 17959 ¶ 98 (mobile providers).

Second, the Order imposes anti-blocking requirements on both types of broadband providers. It prohibits fixed broadband providers from “blockfing] lawful content, applications, services, or non-harmful devices, subject to reasonable network management.” Id. at 17942 ¶ 63. Similarly, the Order forbids mobile providers from “blocking] consumers from accessing lawful websites” and from “blocking] applications that compete with the provider’s voice or video telephony services, subject to reasonable network management.” Id. at 17959 ¶ 99. The Order defines “reasonable network management” as practices designed to “ensur[e] network security and integrity,” “address! ] traffic that is unwanted by end users,” “and reducfe] or mitigatfe] the effects of congestion on the network.” Id. at 17952 ¶ 82. The anti-blocking rules, the Order explains, not only prohibit broadband providers from preventing their end-user subscribers from accessing a particular edge provider altogether, but also prohibit them “from impairing or degrading particular content, applications, services, or non-harmful devices so as to render them effectively unusable.” Id. at 17943 ¶ 66.

Third, the Order imposes an anti-discrimination requirement on fixed broadband providers only. Under this rule, such providers “shall not unreasonably discriminate in transmitting lawful network traffic over a consumer’s broadband Internet access service. Reasonable network management shall not constitute unreasonable discrimination.” Id. at 17944 ¶ 68. The Commission explained that “[u]se-agnostic discrimination” — that is, discrimination based not on the nature of the particular traffic involved, but rather, for example, on network management needs during periods of congestion— would generally comport with this requirement. Id. at 17945-46 ¶ 73. Although the Commission never expressly said that the rule forbids broadband providers from granting preferred status or services to edge providers who pay for such benefits, it warned that “as a general matter, it is unlikely that pay for priority would satisfy the ‘no unreasonable discrimination’ standard.” Id. at 17947 ¶ 76. Declining to impose the same anti-discrimination requirement on mobile providers, the Commission explained that differential treatment of such providers was warranted because the mobile broadband market was more competitive ■ and more rapidly evolving than the fixed broadband market, network speeds and penetration were lower, and operational constraints were higher. See id. at 17956-57 ¶¶ 94-95.

As authority for the adoption of these rules, the Commission invoked a plethora of statutory provisions. See id. at 17966-81 ¶¶ 115-37. In particular, the Commission relied on section 706 of the 1996 Telecommunications Act, which directs it to encourage the deployment of broadband telecommunications capability. See 47 U.S.C. § 1302(a), (b). According to the Commission, the rules furthered this statutory mandate by preserving unhindered the “virtuous circle of innovation” that had long driven the growth of the Internet. Open Internet Order, 25 F.C.C.R. at 17910-11 ¶ 14; see id. at 17968, 17972 ¶¶ 117, 123. Internet openness, it reasoned, spurs investment and development by edge providers, which leads to increased end-user demand for broadband access, which leads to increased investment in broadband network infrastructure and technologies, which in turn leads to further innovation and development by edge providers. Id. at 17910-11 ¶ 14. If, the Commission continued, broadband providers were to disrupt this “virtuous circle” by “[rjestricting edge providers’ ability to reach end users, and limiting end users’ ability to choose which edge providers to patronize,” they would “reduce the rate of innovation at the edge and, in turn, the likely rate of improvements to network infrastructure.” Id. at 17911 ¶ 14.

Two members of the Commission dissented. As they saw it, the Open Internet Order rules not only exceeded the Commission’s lawful authority, but would also stifle rather than encourage innovation. See Open Internet Order, 25 F.C.C.R. at 18049-81 (Dissenting Statement of Commissioner McDowell); id. at 18084-98 (Dissenting Statement of Commissioner Baker).

Verizon filed a petition for review of the Open Internet Order pursuant to 47 U.S.C. § 402(a) as well as a notice of appeal pursuant to 47 U.S.C. § 402(b). Because “we plainly have jurisdiction by the one procedural route or the other,” “we need not decide which is the more appropriate vehicle for our review.” Cellco Partnership v. FCC, 700 F.3d 534, 541 (D.C.Cir.2012) (internal quotation marks omitted).

Verizon challenges the Open Internet Order on several grounds, including that the Commission lacked affirmative statutory authority to promulgate the rules, that its decision to impose the rules was arbitrary and capricious, and that the rules contravene statutory provisions prohibiting the Commission from treating broadband providers as common carriers. In Part II, we consider Verizon’s attacks on the Commission’s affirmative statutory authority and its justification for imposing these rules. We consider the common carrier issue in Part III. Given our disposition of the latter issue, we have no need to address Verizon’s additional contentions that the Order violates the First Amendment and constitutes an uncompensated taking.

Before beginning our analysis, we think it important to emphasize that although the question of net neutrality implicates serious policy questions, which have engaged lawmakers, regulators, businesses, and other members of the public for years, our inquiry here is relatively limited. “Regardless of how serious the problem an administrative agency seeks to address, ... it may not exercise its authority in a manner that is inconsistent with the administrative structure that Congress enacted into law.” Ragsdale v. Wolverine World Wide, Inc., 535 U.S. 81, 91, 122 S.Ct. 1155, 152 L.Ed.2d 167 (2002) (internal quotation marks omitted). Accordingly, our task as a reviewing court is not to assess the wisdom of the Open Internet Order regulations, but rather to determine whether the Commission has demonstrated that the regulations fall within the scope of its statutory grant of authority.

II.

The Commission cites numerous statutory provisions it claims grant it the power to promulgate the Open Internet Order rules. But we start and end our analysis with section 706 of the 1996 Telecommunications Act, which, as we shall explain, furnishes the Commission with the requisite affirmative authority to adopt the regulations.

Section 706(a) provides:

The Commission and each State commission with regulatory jurisdiction over telecommunications services shall encourage the deployment on a reasonable and timely basis of advanced telecommunications capability to all Americans (including, in particular, elementary and secondary schools and classrooms) by utilizing, in a manner consistent with the public interest, convenience, and necessity, price cap regulation, regulatory forbearance, measures that promote competition in the local telecommunications market, or other regulating methods that remove barriers to infrastructure investment.

47 U.S.C. § 1302(a). Section 706(b), in turn, requires the Commission to conduct a regular inquiry “concerning the availability of advanced telecommunications capability.” Id. § 1302(b). It further provides that should the Commission find that “advanced telecommunications capability is [not] being deployed to all Americans in a reasonable and timely fashion,” it “shall take immediate action to accelerate deployment of such capability by removing barriers to infrastructure investment and by promoting competition in the telecommunications market.” Id. The statute defines “advanced telecommunications capability” to include “broadband telecommunications capability.” Id. § 1302(d)(1).

Verizon contends that neither subsection (a) nor (b) of section 706 confers any regulatory authority on the Commission. As Verizon sees it, the two subsections amount to nothing more than congressional statements of policy. Verizon further contends that even if either provision grants the Commission substantive authority, the scope of that grant is not so expansive as to permit the Commission to regulate broadband providers in the manner that the Open Internet Order rules do. In addressing these questions, we apply the familiar two-step analysis of Chevron, U.S.A., Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984). As the Supreme Court has recently made clear, Chevron deference is warranted even if the Commission has interpreted a statutory provision that could be said to delineate the scope of the agency’s jurisdiction. See City of Arlington v. FCC,—U.S.-, 133 S.Ct. 1863, 1874,—L.Ed.2d - (2013). Thus, if we determine that the Commission’s interpretation of section 706 represents a reasonable resolution of a statutory ambiguity, we must defer to that interpretation. See Chevron, 467 U.S. at 842-43, 104 S.Ct. 2778. The Chevron inquiry overlaps substantially with that required by the Administrative Procedure Act (APA), pursuant to which we must also determine whether the Commission’s actions were “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.” 5 U.S.C. § 706(2)(A); see National Ass’n of Regulatory Utility Commissioners v. Interstate Commerce Commission, 41 F.3d 721, 726-27 (D.C.Cir.1994).

A.

This is not the first time the Commission has asserted that section 706(a) grants it authority to regulate broadband providers. Advancing a similar argument in Comcast, the Commission contended that section 706(a) provided a statutory hook for its exercise of ancillary jurisdiction. Although we thought that section 706(a) might “arguably be read to delegate regulatory authority to the Commission,” we concluded that the Commission could not rely on this provision to justify the Comcast Order because it had previously determined, in the still-binding Advanced Services Order, that the provision “‘does not constitute an independent grant of authority.’ ” Comcast, 600 F.3d at 658 (quoting Advanced Services Order, 13 F.C.C.R. at 24047 ¶ 77). We rejected the Commission’s claim that the Advanced Services Order concluded only that section 706(a) granted it no forbearance authority — authority to relieve regulated entities of statutory obligations to which they would otherwise be subject, see 47 U.S.C. § 160— over and above that given it elsewhere in the Communications Act. Comcast, 600 F.3d at 658. Indeed, the Advanced Services Order was clearly far broader, explicitly declaring: “section 706(a) does not constitute an independent grant of forbearance authority or of authority to employ other regulating methods.” Advanced Services Order, 13 F.C.C.R. at 24044 ¶ 69 (emphasis added). Because the Commission had “never questioned, let alone overruled, that understanding of section 706,” we held that it “remain[ed] bound” by its prior interpretation. Com-cast, 600 F.3d at 659.

But the Commission need not remain forever bound by the Advanced Services Order’s restrictive reading of section 706(a). “An initial agency interpretation is not instantly carved in stone.” Chevron, 467 U.S. at 863, 104 S.Ct. 2778. The APA’s requirement of reasoned decision-making ordinarily demands that an agency acknowledge and explain the reasons for a changed interpretation. See FCC v. Fox Television Stations, Inc., 556 U.S. 502, 515, 129 S.Ct. 1800, 173 L.Ed.2d 738 (2009) (“An agency may not ... depart from a prior policy sub silentio or simply disregard rules that are still on the books.”); Brand X, 545 U.S. at 981, 125 S.Ct. 2688 (“Unexplained inconsistency is, at most, a reason for holding an interpretation to be an arbitrary and capricious change from agency practice under the Administrative Procedure Act.”). But so long as an agency “adequately explains the reasons for a reversal of policy,” its new interpretation of a statute cannot be rejected simply because it is new. Brand X, 545 U.S. at 981, 125 S.Ct. 2688. At the time we issued our Comcast opinion, the Commission failed to satisfy this requirement, as its assertion that section 706(a) gave it regulatory authority represented, at that point, an attempt to “ ‘depart from a prior policy sub silentio.’ ” Comcast, 600 F.3d at 659 (quoting Fox, 556 U.S. at 515, 129 S.Ct. 1800).

In the Open Internet Order, however, the Commission has offered a reasoned explanation for its changed understanding of section 706(a).’ To be sure, the Open Internet Order evinces a palpable reluctance to accept this court’s interpretation of the Advanced Services Order, as the Commission again attempts to reconcile its current understanding of section 706(a) with its prior interpretation. See Open Internet Order, 25 F.C.C.R. at 17969 ¶ 119 (characterizing the Advanced Services Order as being “consistent with [the Commission’s] present understanding”). Of course, such reluctance hardly makes the Commission’s decision unreasonable, as it is free to express its disagreement with this court’s holdings. After all, even a federal agency is entitled to a little pride. Moreover, although the Open Internet Order inaccurately describes the Advanced Services Order’s actual conclusion, it does describe what the Order likely should have concluded. Specifically, the Advanced Services Order’s rejection of section 706(a) as a source of substantive authority rested almost entirely on the notion that a contrary interpretation would somehow permit the Commission to evade express statutory commands forbidding it from using its forbearance authority in certain circumstances. See Advanced Services Order, 13 F.C.C.R. at 24045-46 ¶¶ 72-73. This makes little sense. By the same reasoning, one might say that Article I of the Constitution gives Congress no substantive authority because Congress might otherwise be able to use that authority in a way that violates the Ex Post Facto Clause. The Open Internet Order characterizes the Advanced Services Order as simply “disavowing a reading of Section 706(a) that would allow the agency to trump specific mandates of the Communications Act,” thus honoring “the interpretive canon that ‘[a] specific provision ... controls one[ ] of more general application.’ ” Open Internet Order, 25 F.C.C.R. at 17969 ¶¶ 118-119 (quoting Bloate v. United States, 559 U.S. 196, 207, 130 S.Ct. 1345, 176 L.Ed.2d 54 (2010)). Perhaps the Commission should have more openly acknowledged that it was not actually describing the Advanced Services Order, but instead rewriting it in a more logical manner. In this latter task, however, the Commission succeeded: its reinterpretation of the Advanced Services Order was more reasonable than the Advanced Services Order itself.

In any event — and more important for our purposes — the Commission expressly declared: “To the extent that the Advanced Services Order can be construed as having read Section 706(a) differently, we reject that reading of the statute for the reasons discussed in the text.” Open Internet Order, 25 F.C.C.R. at 17969 ¶ 119 n. 370. Setting forth those “reasons” at some length, the Commission analyzed the statute’s text, its legislative history, and the resultant scope "of the Commission’s authority, concluding that each of these considerations supports the view that section 706(a) constitutes an affirmative grant of regulatory authority. Id. at 17969-70 ¶¶ 119-121. In these circumstances, and contrary to Verizon’s contentions, we have no basis for saying that the Commission “casually ignored prior policies and interpretations or otherwise failed to provide a reasoned explanation” for its changed interpretation. Cablevision Systems Corp. v. FCC, 649 F.3d 695, 710 (D.C.Cir.2011) (internal quotation marks omitted).

The question, then, is this: Does the Commission’s current understanding of section 706(a) as a grant of regulatory authority represent a reasonable interpretation of an ambiguous statute? We believe it does.

Recall that the provision directs the Commission to “encourage the deployment ... of advanced telecommunications capability ... by utilizing ... price cap regulation, regulatory forbearance, measures that promote competition in the local telecommunications market, or other regulating methods that remove barriers to infrastructure investment.” 47 U.S.C. § 1302(a). As Verizon argues, this language could certainly be read as simply setting forth a statement of congressional policy, directing the Commission to employ “regulating methods” already at the Commission’s disposal in order to achieve the stated goal of promoting “advanced telecommunications” technology. But the language can just as easily be read to vest the Commission with actual authority to utilize such “regulating methods” to meet this stated goal. As the Commission put it in the Open Internet Order, one might reasonably think that Congress, in directing the Commission to undertake certain acts, “necessarily invested the Commission with the statutory authority to carry out those acts.” Open Internet Order, 25 F.C.C.R. at 17969 ¶ 120.

Section 706(a)’s reference to state commissions does not foreclose such a reading. Observing that the statute applies to both “[t]he Commission and each State commission with regulatory jurisdiction over telecommunications services,” 47 U.S.C. § 1302(a) (emphasis added), Verizon contends that Congress would not be expected to grant both the FCC and state commissions the regulatory authority to encourage the deployment of advanced telecommunications capabilities. But Congress has granted regulatory authority to state telecommunications commissions on other occasions, and we see no reason to think that it could not have done the same here. See, e.g., id. § 251(f) (granting state commissions the authority to exempt rural local exchange carriers from certain obligations imposed on other incumbents); id. § 252(e) (requiring all interconnection agreements between incumbent local exchange carriers and entrant carriers to be approved by a state commission); see also AT & T Corp. v. Iowa Utilities Board, 525 U.S. 366, 385-86, 119 S.Ct. 721, 142 L.Ed.2d 835 (1999) (describing the Commission’s power and responsibility to dictate the manner in which state commissions exercise such authority). Thus, Congress has not “directly spoken” to the question of whether section 706(a) is a grant of regulatory authority simply by mentioning state commissions in that grant. Chevron, 467 U.S. at 842, 104 S.Ct. 2778.

This case, moreover, is a .far cry from FDA v. Brown & Williamson Tobacco Corp., 529 U.S. 120, 120 S.Ct. 1291, 146 L.Ed.2d 121 (2000), on which Verizon principally relies. There, the Supreme Court held that “Congress ha[d] clearly precluded the [Food and Drug Administration] from asserting jurisdiction to regulate tobacco products.” Id. at 126, 120 S.Ct. 1-291. The Court emphasized that the FDA had not only completely disclaimed any authority to regulate tobacco products, but had done so for more than eighty years, and that Congress had repeatedly legislated against this background. See id. at 143-59, 120 S.Ct. 1291. The Court also observed that the FDA’s newly adopted conclusion that it did in fact have authority to regulate this industry would, given its findings regarding the .effects of tobacco products and its authorizing statute, logically require the agency to ban such products altogether,'a result clearly contrary to congressional policy. See id. at 135-43, 120 S.Ct. 1291. Furthermore, the Court reasoned, if Congress had intended to “delegate a decision of such economic and political significance” to the agency, it would have done so far more clearly. Id. at 160, 120 S.Ct. 1291.

The circumstances here are entirely different. Although the Commission once disclaimed authority to regulate under section 706(a), it never disclaimed authority to regulate the Internet or Internet providers altogether, nor is there any similar history of congressional reliance on such a disclaimer. To the contrary, as recounted above, see supra at 629-31, when Congress passed section 706(a) in 1996, it did so against the backdrop of the Commission’s long history of subjecting to common carrier regulation the entities that controlled the last-mile facilities over which end users accessed the Internet. See, e.g., Second Computer Inquiry, 77 F.C.C.2d at 473-74 ¶¶ 228-29. Indeed, one might have thought, as the Commission originally concluded, see Advanced Services Order, 13 F.C.C.R. at 24029-30 ¶ 35, that Congress clearly contemplated that the Commission would continue regulating Internet providers in the manner it had previously. Cf. Brand X, 545 U.S. at 1003, 125 S.Ct. 2688 (Breyer, J., concurring) (concluding that the Commission’s decision to exempt cable broadband providers from Title II regulation was “perhaps just barely” within the scope of the agency’s “statutorily delegated authority”); id. at 1005, 125 S.Ct. 2688 (Scalia, J., dissenting) (arguing that Commission’s decision “exceeded the authority given it by Congress”). In fact, section 706(a)’s legislative history suggests’ that Congress may have, somewhat presciently, viewed that provision as an affirmative grant of authority to the Commission whose existence would become necessary if other contemplated grants of statutory authority were for some reason unavailable. The Senate Report describes section 706 as a “necessary fail-safe” “intended to ensure that one of the primary objectives of the [Act] — to accelerate deployment of advanced telecommunications capability — is achieved.” S.Rep. No. 104-23 at 50-51. As the Commission observed in the Open Internet Order, it would be “odd ... to characterize Section 706(a) as a ‘fail-safe’ that ‘ensures’ the Commission’s ability to promote advanced services if it conferred no actual authority.” 25 F.C.C.R. at 17970 ¶ 120.

Verizon directs our attention to a number of bills introduced in Congress subsequent to the passage of the 1996 Act that, if enacted, would have imposed requirements on broadband providers similar to those embodied in the Commission’s Open Internet Order. See, e.g., Internet Non-Discrimination Act of 2006, S. 2360, 109th Cong. (2006). Such subsequent legislative history, however, provides “ ‘an unreliable guide to legislative intent.’ ” North Broward Hospital District v. Shala-la, 172 F.3d 90, 98 (D.C.Cir.1999) (quoting Chapman v. United States, 500 U.S. 453, 464 n. 4, 111 S.Ct. 1919, 114 L.Ed.2d 524 (1991)). Moreover, even assuming that Congress’s failure to impose such restrictions would itself cast light on Congress’s understanding of the Commission’s power to do so, any such inferences would be largely countered by Congress’s similar failure to adopt a proposed resolution that would have specifically disapproved of the Commission’s promulgation of the Open Internet Order. See H.J. Res. 37, 112th Cong. (2011). These conflicting pieces of subsequent failed legislation tell us. little if anything about the original meaning of the Telecommunications Act of 1996.

Thus, although regulation of broadband Internet providers certainly involves decisions of great “economic and political significance,” Brown & Williamson, 529 U.S. at 160, 120 S.Ct. 1291, we have little reason given this history to think that Congress could not have delegated some of these decisions to the Commission. To be sure, Congress does not, as Verizon reminds us, “hide elephants in mouseholes.” Whitman v. American Trucking Ass’ns, Inc., 531 U.S. 457, 468, 121 S.Ct. 903, 149 L.Ed.2d 1 (2001). But FCC regulation of broadband providers is no elephant, and section 706(a) is no mousehole.

Of course, we might well hesitate to conclude that Congress intended to grant the Commission substantive authority in section 706(a) if that authority would have no limiting principle. See Comcast, 600 F.3d at 655 (rejecting Commission’s understanding of its authority that “if accepted ... would virtually free the Commission from its congressional tether”); cf. Whitman, 531 U.S. at 472-73, 121 S.Ct. 903 (discussing the nondelegation doctrine). But we are satisfied that the scope of authority granted to the Commission by section 706(a) is not so boundless as to compel the conclusion that Congress could never have intended the provision to set forth anything other than a general statement of policy. The Commission has identified at least two limiting principles inherent in section 706(a). See Open Internet Order, 25 F.C.C.R. at 17970 ¶ 121. First, the section must be read in conjunction with other provisions of the Communications Act, including, most importantly, those limiting the Commission’s subject matter jurisdiction to “interstate and foreign communication by wire and radio.” 47 U.S.C. § 152(a). Any regulatory action authorized by section 706(a) would thus have to fall within the Commission’s subject matter jurisdiction over such communications — a limitation whose importance this court has recognized in delineating the reach of the Commission’s ancillary jurisdiction. See American Library Ass’n, 406 F.3d at 703-04. Second, any regulations must be designed to achieve a particular purpose: to “encourage the deployment on a reasonable and timely basis of advanced telecommunications capability to all Americans.” 47 U.S.C. § 1302(a). Section 706(a) thus gives the Commission authority to promulgate only those regulations that it establishes will fulfill this specific statutory goal — a burden that, as we trust our searching analysis below will demonstrate, is far from “meaningless.” Dissenting Op. at 662.

B.

Section 706(b) has a less tortured history. Until shortly before the Commission issued the Open Internet Order, it had never considered whether the provision vested it with any regulatory authority. The Commission had no need to do so because prior to that time it had made no determination that advanced telecommunications technologies, including broadband Internet access, were not “being deployed to all Americans in a reasonable and timely fashion,” the prerequisite for any purported invocation of authority to “take immediate action to accelerate deployment of such capability” under section 706(b). 47 U.S.C. § 1302(b).

In July 2010, however, the Commission concluded that “broadband deployment to all Americans is not reasonable and timely.” Sixth Broadband Deployment Report, 25 F.C.C.R. at 9558 ¶2. This conclusion, the Commission recognized, represented a deviation from its five pri- or assessments. Id. at 9558 ¶2 & n. 8. According to the Commission, the change was driven by its decision to raise the minimum speed threshold qualifying as broadband. Id. at 9558 ¶ 4. “Broadband,” as defined in the 1996 Telecommunications Act, is Internet service furnished at speeds that “enable[ ] users to originate and receive high-quality voice, data, graphics, and video telecommunications using any technology.” 47 U.S.C. § 1302(d)(1). In 1999, the Commission found this requirement satisfied by services “having the capability of supporting ... a speed ... in excess of 200 kilobits per second (kbps) in the last mile.” In re Inquiry Concerning the Deployment of Advanced Telecommunications Capability to All Americans in a Reasonable and Timely Fashion, 14 F.C.C.R. 2398, 2406 ¶ 20 (1999). The Commission chose this threshold because it was “enough to provide the most popular forms of broadband — to change web pages as fast as one can flip through the pages of a book and to transmit full-motion video.” Id. That said, the Commission recognized that technological developments might someday require it to reassess the 200 kbps threshold. Id. at 2407-08 ¶ 25.

In the Sixth Broadband Deployment Report, the Commission decided that day had finally arrived. The Commission explained that consumers now regularly use their Internet connections to access high-quality video and expect to be able at the same time to check their email and browse the web. Sixth Broadband Deployment Report, 25 F.C.C.R. at 9562-64 ¶¶ 10-11. Two hundred kbps, the Commission determined, “simply is not enough bandwidth” to permit such uses. Id. at 9562 ¶ 10. The Commission thus adopted a new threshold more appropriate to current consumer behavior and expectations: four megabytes per. second (mbps) for end users to download content from the Internet — twenty times as fast as the prior threshold — and one mbps for end users to upload content. Id. at 9563 ¶ 11.

Applying this new benchmark, the Commission found that “roughly 80 million American adults do not subscribe to broadband at home, and approximately 14 to 24 million Americans do not have access to broadband today.” Sixth Broadband Deployment Report, 25 F.C.C.R. at 9574 ¶ 28. Given these figures and the “ever-growing importance of broadband to our society,” the Commission was unable to find “that broadband is being reasonably and timely deployed” within the meaning of section 706(b). Id. This conclusion, it explained; triggered section 706(b)’s mandate that the Commission “take immediate action to accelerate deployment.” Id. at 9558 ¶3 (quoting 47 U.S.C. § 1302(b)) (internal quotation marks omitted).

Subsequently, in the Open Internet Order the Commission made clear that this statutory provision does not limit the Commission to using other regulatory authority already at its disposal, but instead grants it the power necessary to fulfill the statute’s mandate. See Open Internet Order, 25 F.C.C.R. at 17972 ¶ 123. Emphasizing the provision’s “shall take immediate action” directive, the Commission concluded that section 706(b) “provides express authority” for the rules it adopted. Id.

Contrary to Verizon’s arguments, we believe the Commission has reasonably interpreted section 706(b) to empower it to take steps to accelerate broadband deployment if and when it determines that such deployment is not “reasonable and timely.” To be sure, as with section 706(a), it is unclear whether section 706(b), in providing that the Commission “shall take immediate action to accelerate deployment of such capability by removing barriers to infrastructure investment and by promoting competition in the telecommunications market,” vested the Commission with authority to remove such barriers to infrastructure investment and promote competition. 47 U.S.C. § 1302(b), But the provision may certainly be read to accomplish as much, and given such ambiguity we have no basis for rejecting the Commission’s determination that it should be so understood. See Chevron, 467 U.S. at 842-43, 104 S.Ct. 2778. Moreover, as discussed above with respect to section 706(a), see supra at 638-40, nothing in the regulatory background or the legislative history either before or after passage of the 1996 Telecommunications Act forecloses such an understanding. We think it quite reasonable to believe that Congress contemplated that the Commission would regulate this industry, as the agency had in the past, and the scope of any authority granted to it by section 706(b) — limited, as it is, both by the boundaries of the Commission’s subject matter jurisdiction and the requirement that any regulation be tailored to the specific statutory goal of accelerating broadband deployment — is not. so broad that we might hesitate to think that Congress could have intended such a delegation.

Verizon makes two additional arguments regarding the Commission’s interpretation of section 706(b), both of which we can dispose of in relatively short order.

First, Verizon contends that if section 706(b) gives the Commission any regulatory authority, that authority must be understood in conjunction with section 706(c), which directs the Commission to “compile a list of geographical areas that are not served by any provider of advanced telecommunications capability.” 47 U.S.C. § 1302(c). Thus, Verizon claims, any regulations that the Commission might adopt pursuant to section 706(b) may not “reach beyond any particular ‘geographical areas that are not served’ by any broadband provider and apply throughout the country.” Verizon’s Br. 33 (emphasis omitted). By its own terms, however, section 706(c) describes simply “part of the inquiry” that section 706(b) requires the Commission to conduct concerning broadband deployment. 47 U.S.C. § 1302(c) (emphasis added). It nowhere purports to delineate all aspects of that inquiry. Nor does it limit the actions that the Commission may take if, in the course of that inquiry, it determines that broadband deployment has not been “reasonable and timely.”

Second, Verizon asserts that the Sixth Broadband Deployment Report’s, finding that triggered section 706(b)’s grant of regulatory authority “arbitrarily contravened five prior agency determinations of reasonable and timely deployment.” Verizon’s Br. 33. The timing of the Commission’s determination is certainly suspicious, coming as it did closely on the heels of our rejection in Comcast of the legal theory on which the Commission had until then relied to establish its authority over broadband providers. But questionable timing, by itself, gives us no basis to reject an otherwise reasonable finding. Beyond its general assertion that the Commission’s finding was “arbitrarily],” Verizon offers no specific reason for thinking that the Commission’s logical and carefully reasoned determination was illegitimate. We can see none.

C.

This brings us, then, to Verizon’s alternative argument that even if, as we have held, sections 706(a) and 706(b) grant the Commission affirmative authority to promulgate rules governing broadband providers, the specific rules imposed by the Open Internet Order fall outside the scope of that authority. The Commission’s theory, to reiterate, is that its regulations protect and promote edge-provider investment and development, which in turn drives end-user demand for more and better broadband technologies, which in turn stimulates competition among broadband providers to further invest in broadband. See Open Internet Order, 25 F.C.C.R. at 17910-11, 17970 ¶¶ 14, 120. Thus, the Commission claims, by preventing broadband providers from blocking or discriminating against edge providers, the rules “encourage the deployment on a reasonable and timely basis of advanced telecommunications capability to all Americans,” 47 U.S.C. § 1302(a), and “accelerate deployment of such capability,” id. § 1302(b), by removing “barriers to infrastructure investment” and promoting “competition,” id. § 1302(a), (b). See Open Internet Order, 25 F.C.C.R. at 17968, 17972 ¶¶117, 123. That is, contrary to the dissent, see Dissenting Op. at 660-62, the Commission made clear — and Verizon appears to recognize — that the Commission found broadband providers’ potential disruption of edge-provider traffic to be itself the sort of “barrier” that has “the potential to stifle overall investment in Internet infrastructure,” and could “limit competition in telecommunications markets.” Open Internet Order, 25 F.C.C.R. at 17970 ¶ 120.

Verizon mounts a twofold challenge to this rationale. It argues that the Open Internet Order regulations will not, as the Commission claims, meaningfully promote broadband deployment, and that even if they do advance this goal, the manner in which they do so is too attenuated from this statutory purpose to fall within the scope of authority granted by either statutory provision.

We begin with the second, more strictly legal, question of whether, assuming the Commission has accurately predicted the effect of these regulations, it may utilize the authority granted to it in sections 706(a) and 706(b) to impose regulations of this sort on broadband providers. As we have previously acknowledged, “in proscribing ... practices with the statutorily identified effect, an agency might stray so far from the paradigm case as to render its interpretation unreasonable, arbitrary, or capricious.” National Cable & Telecommunications Ass’n v. FCC, 567 F.3d 659, 665 (D.C.Cir.2009). Here, Verizon has given us no reason to conclude that the Open Internet Order's requirements “stray” so far beyond the “paradigm case” that Congress likely contemplated as to render the Commission’s understanding of its authority unreasonable. The rules not only apply directly to broadband providers, the precise entities to which section 706 authority to encourage broadband deployment presumably extends, but also seek to promote the very goal that Congress explicitly sought to promote. Because the rules advance this statutory goal of broadband deployment by first promoting edge-provider innovations and end-user demand, Verizon derides the Commission’s justification as a “triple-cushion shot.” Verizon’s Br. 28. In billiards, however, a triple-cushion shot, although perhaps more difficult to complete, counts the same as any other shot. The Commission could reasonably have thought that its authority to promulgate regulations that promote broadband deployment encompasses the power to regulate broadband providers’ economic relationships with edge providers if, in fact, the nature of those relationships influences the rate and extent to which broadband providers develop and expand their services for end users. See Cablevision, 649 F.3d at 709 (holding that Commission had not impermissibly “reached beyond the paradigm case” in “interpreting a statute focused on the provision of satellite programming to authorize terrestrial withholding regulations,” because cable companies’ ability to withhold terrestrial programming would, in turn, discourage potential competitors from entering the market to provide satellite programming) (internal quotation marks omitted).

Whether the Commission’s assessment of the likely effects of the Open Internet Order deserves credence presents a slightly more complex question. Verizon attacks the reasoning and factual support underlying the Commission’s “triple-cushion shot” theory, advancing these arguments both as an attack on the Commission’s statutory interpretation and as an APA arbitrary and capricious challenge. Given that these two arguments involve similar considerations, we address them together. In so doing, “we must uphold the Commission’s factual determinations if on the record as a whole, there is such relevant evidence as a reasonable mind might accept as adequate to support [the] conclusion.” Secretary of Labor, MSHA v. Federal Mine Safety & Health Review Comm’n, 111 F.3d 913, 918 (D.C.Cir.1997) (internal quotation marks omitted); see 5 U.S.C. § 706(2)(E). We evaluate the Commission’s reasoning to ensure that it has “examine[d] the relevant data and articulate[d] a satisfactory explanation for its action including a rational connection between the facts found and the choice made.” National Fuel Gas Supply Corp. v. FERC, 468 F.3d 831, 839 (D.C.Cir.2006) (quoting Motor Vehicle Manufacturers Ass’n of U.S. v. State Farm Mutual Auto. Insurance Co., 463 U.S. 29, 43, 103 S.Ct. 2856, 77 L.Ed.2d 443 (1983)) (internal quotation marks omitted). When assessing the reasonableness of the Commission’s conclusions, we must be careful not to simply “ ‘substitute [our] judgment for that of the agency,’ ” especially when the “agency’s predictive judgments about the likely economic effects of a rule” are at issue. National Telephone Cooperative Ass’n v. FCC, 563 F.3d 536, 541 (D.C.Cir.2009) (quoting State Farm, 463 U.S. at 43, 103 S.Ct. 2856). Under these standards, the Commission’s prediction that the Open Internet Order regulations will encourage broadband deployment is, in our view, both rational and supported by substantial evidence.

To begin with, the Commission has more than adequately supported and explained its conclusion that edge-provider innovation leads to the expansion and improvement of broadband infrastructure. The Internet, the Commission observed in the Open Internet Order, is, “[l]ike electricity and the computer,” a “ ‘general purpose technology’ that enables new methods of production that have a major impact on the entire economy.” Open Internet Order, 25 F.C.C.R. at 17909 ¶ 13. Certain innovations — the lightbulb, for example— create a need for infrastructure investment, such as in power generation facilities and distribution lines, that complement and further drive the development of the initial innovation and ultimately the growth of the economy as a whole. See Timothy F. Bresnahan & M. Trajtenberg, General purpose technologies: ‘Engines of Growth’? 65 J. Econometrics 83, 84 (1995), cited in Open Internet Order, 25 F.C.C.R. at 17909 ¶ 13 n. 12; see also Amicus Br. of Internet Engineers and Technologists 17 (citing Hearing on Internet Security Before the H. Comm, on Science, Space, and Technology, 103d Cong. (Mar. 22, 1994) (written testimony of Dr. Vinton G. Cerf)). The rise of streaming online video is perhaps the best and clearest example the Commission used to illustrate that the Internet constitutes one such technology: higher-speed residential Internet connections in the late 1990s “stimulated” the development of streaming video, a service that requires particularly high bandwidth, “which in turn encouraged broadband providers to increase network speeds.” Open Internet Order, 25 F.C.C.R. at 17911 ¶ 14 n. 23. The Commission’s emphasis on this connection between edge-provider innovation and infrastructure development is uncontroversial. Indeed, in its comments to the Commission, Verizon, executing a triple-cushion shot of its own, acknowledged:

[T]he social and economic fruits of the Internet economy are the result of a virtuous cycle of'innovation and growth between that ecosystem and the underlying infrastructure — the infrastructure enabling the development and dissemination of Internet-based services and applications, with the demand and use of those services ... driving improvements in the infrastructure which, in turn, support further innovations -in services and applications.

Verizon Comments at 42, Docket No. 09-191 (Jan. 14, 2010) (internal quotation marks omitted).

The Commission’s finding that Internet openness fosters the edge-provider innovation that drives this “virtuous cycle” was likewise reasonable and grounded in substantial evidence. Continued innovation at the edge, the Commission explained, “depends upon low barriers to innovation and entry by edge providers,” and thus restrictions on edge providers’ “ability to reach end users ... reduce the rate of innovation.” Open Internet Order, 25 F.C.C.R. at 17911 ¶ 14. This conclusion finds ample support in the economic literature on which the Commission relied, see, e.g., Joseph Farrell & Philip J. Weiser, Modularity, Vertical Integration, and Open Access Policies: Towards a Convergence of Antitrust and Regulation in the Internet Age, 17 Harv. J.L. & Tech. 85, 95 (2003), cited in Open Internet Order, 25 F.C.C.R. at 17911 ¶ 14 n. 25, as well as in history and the comments of several edge providers. For one prominent illustration of the relationship between openness and innovation, the Commission cited the invention of the World Wide Web itself by Sir Tim Berners-Lee, who, although not working for an entity that operated the underlying network, was able to create and disseminate this enormously successful innovation without needing to make any changes to previously developed Internet protocols or securing “any approval from network operators.” Open Internet Order, 25 F.C.C.R. at 17910 ¶ 13 (citing, inter alia, Tim Berners-Lee, Weaving the Web 16 (2000)). It also highlighted the comments of Google and Vonage — both innovative edge providers — who emphasized the importance of the Internet’s open design to permitting new content and services to develop at the edge. Id. at 17911 ¶ 14 n. 24 & n. 25. The record amassed by the Commission contains many similar examples, and Verizon has given us no basis for questioning the Commission’s determination that the preservation of Internet openness is integral to achieving the statutory objectives set forth in Section 706. See id. at 17910-11, 17968, 17972 ¶¶ 14, 117, 123.

Equally important, the Commission has adequately supported and explained its conclusion that, absent rules such as those set forth in the Open Internet Order, broadband providers represent a threat to Internet openness and could act in ways that would ultimately inhibit the speed and extent of future broadband deployment. First, nothing in the record gives us any reason to doubt the Commission’s determination that broadband providers may be motivated to discriminate against and among edge providers. The Commission observed that broadband providers— often the same entities that furnish end users with telephone and television services — “have incentives to interfere with the operation of third-party Internet-based services that compete with the providers’ revenue-generating telephone and/or pay-television services.” Open Internet Order, 25 F.C.C.R. at 17916 ¶ 22. As the Commission noted, Voice-over-Internet-Protocol (VoIP) services such as Vonage increasingly serve as substitutes for traditional telephone services, id., and broadband providers like AT & T and Time Warner have acknowledged that online video aggregators such as Netflix and Hulu compete directly with their own “core video subscription service,” id. at 17917 ¶22 & n. 54; see also id. at 17918 ¶ 23 n. 60 (finding that a study concluding that cable companies had sought to exclude networks that competed with the companies’ own affiliated channels, see Austan Goolsbee, Vertical Integration and the Market for Broadcast and Cable Television Programming, Paper for the Federal Communications Commission 31-32 (Sept. 5, 2007), “provides empirical evidence that cable providers have acted in the past on anticompetitive incentives to foreclose rivals”). Broadband providers also have powerful incentives to accept fees from edge providers, either in return for excluding their competitors or for granting them prioritized access to end users. See id. at 17918-19 ¶¶ 23-24. Indeed, at oral argument Verizon’s counsel announced that “but for [the Open Internet Order ] rules we would be exploring those commercial arrangements.” Oral Arg. Tr. 31. And although broadband providers might not adopt pay-for-priority agreements or other similar arrangements if, according to the Commission’s analysis, such agreements would ultimately lead to a decrease in end-user demand for broadband, the Commission explained that the resultant harms to innovation and demand will largely constitute “negative externalities”: any given broadband provider will “receive the benefits of ... fees but [is] unlikely to fully account for the detrimental impact on edge providers’ ability and incentive to innovate and invest.” Open Internet Order, 25 F.C.C.R. at 17919-20 ¶ 25 & n. 68. Although Verizon dismisses the Commission’s assertions regarding broadband providers’ incentives as “pure speculation,” Verizon’s Br. 52, see also Dissenting Op. at 666-67, those assertions are, at the very least, speculation based firmly in common sense and economic reality.

Moreover, as the Commission found, broadband providers have the technical and economic ability to impose such restrictions. Verizon does not seriously contend otherwise. In fact, there appears little dispute that broadband providers have the technological ability to distinguish between and discriminate against certain types of Internet traffic. See Open Internet Order, 25 F.C.C.R. at 17923 ¶31 (broadband providers possess “increasingly sophisticated network management tools” that enable them to “make fine-grained distinction in their handling of network traffic”). The Commission also convincingly detailed how broadband providers’ position in the market gives them the economic power to restrict edge-provider traffic and charge for the services they furnish edge providers. Because all end users generally access the Internet through a single broadband provider, that provider functions as a “ ‘terminating monopolist,’ ” id. at 17919 ¶ 24 n. 66, with power to act as a “gatekeeper” with respect to edge providers that might seek to reach its end-user subscribers, id. at 17919 ¶24. As the Commission reasonably explained, this ability to act as a “gatekeeper” distinguishes broadband providers from other participants in the Internet marketplace — including prominent and potentially powerful edge providers such as Google and Apple — who have no similar “control [over] access to the Internet for their subscribers and for anyone wishing to reach those subscribers.” Id. at 17935 ¶ 50.

To be sure, if end users could immediately respond to any given broadband provider’s attempt to impose restrictions on edge providers by switching broadband providers, this gatekeeper power might well disappear. Cf. Open Internet Order, 25 F.C.C.R. at 17935 ¶ 51 (declining to impose similar rules on “dial-up Internet access service because telephone service has historically provided the easy ability to switch among competing dial-up Internet access' services”). For example, a broadband provider like Comcast would be unable to threaten Netflix that it would slow Netflix traffic if all Comcast subscribers would then immediately switch to a competing broadband provider. But we see no basis for questioning the Commission’s conclusion that end users are unlikely to react in this fashion. According to the Commission, “end users may not know whether charges or service levels their broadband provider is imposing on edge providers vary from those of alternative broadband providers, and even if they do have this information may find it costly to switch.” Id. at 17921 ¶ 27. As described by numerous commenters, and detailed more thoroughly in a Commission report compiling the results of an extensive consumer survey, the costs of switching include: “early termination fees; the inconvenience of ordering, installation, and setup, and associated deposits or fees; possible difficulty returning the earlier broadband provider’s equipment and the cost of replacing incompatible customer-owned equipment; the risk of temporarily losing service; the risk of problems learning how to use the new service; and the possible loss of a provider-specific email address or website.” Open Internet Order, 25 F.C.C.R. at 17924-25 ¶ 34 (footnotes omitted) (citing, inter alia, Federal Communications Commission, Broadband Decisions: What Drives Consumers to Switch — Or Stick With — Their Broadband Internet Provider (FCC Working Paper, Dec. 2010), available at hraunfoss.fcc.gov/ edocs — publie/attachmatch/DOC-303264Al.pdf). Moreover, the Commission emphasized, many end users may have no option to switch, or at least face very limited options: “[a]s of December 2009, nearly 70 percent of households lived in census tracts where only one or two wireline or fixed wireless firms provided” broadband service. Id. at 17923 ¶ 32. As the Commission concluded, any market power that such broadband providers might have with respect to end users would only increase their power with respect to edge providers. Id.

The dissent focuses on this latter aspect of the Commission’s reasoning, arguing at some length that the Commission’s failure to expressly find that broadband providers have market power with respect to end users is “fatal to its attempt to regulate.” Dissenting Op. at 665. But Verizon has never argued that the Commission’s failure to make a market power finding somehow rendered its understanding of its statutory authority unreasonable or its decision arbitrary and capricious. Verizon does fleetingly mention the market power issue once in its opening brief, asserting as part of its First Amendment claim that Turner Broadcasting System, Inc. v. FCC, 520 U.S. 180, 117 S.Ct. 1174, 137 L.Ed.2d 369 (1997) — in which the Supreme Court, applying intermediate scrutiny, upheld a congressional statute compelling cable companies to carry local broadcast television stations, id. at 185, 117 S.Ct. 1174 — is distinguishable in part because, unlike the Commission here, Congress had found “evidence of ‘considerable and growing market power.’” Verizon Br. 46 (quoting Turner, 520 U.S. at 197, 117 S.Ct. 1174). But to say, as Verizon does, that an allegedly speech-infringing regulation violates the First Amendment because of the absence of a market condition that would increase the need for that regulation is hardly to say that the absence of this market condition renders the regulation wholly irrational. Verizon’s bare citation to a Justice Department submission — relied upon by the dissent, see Dissenting Op. at 664, 666-67 — is even less on point, as that submission simply advised the Commission to take care to avoid stifling incentives for broadband investment; it never asserted, as the dissent does, that such market power is required for broadband providers to have the economic clout to restrict edge-provider traffic in the first place. See Department of Justice Comments at 28, Docket No. 09-51 (Jan. 14, 2010). Indeed, when pressed at oral argument to embrace our dissenting colleague’s position, Verizon’s counsel failed to do so, stating only that it was “possible” that if the Commission had made a market power finding, the Order could be justified. Oral Arg. Tr. 10. As we “do not sit as [a] self-directed board[ ] of legal inquiry and research,” and Verizon “has made no attempt to address the issue,” the argument is clearly forfeited. Car-ducci v. Regan, 714 F.2d 171, 177 (D.C.Cir.1983).

In any event, it seems likely that the reason Verizon never advanced this argument is that the Commission’s failure to find market power is not “fatal” to its theory. Broadband providers’ ability to impose restrictions on edge providers does not depend on their benefiting from the sort of market concentration that would enable them to impose substantial price increases on end users — which is all the Commission said in declining to make a market power finding. See Open Internet Order, 25 F.C.C.R. at 17923 ¶32 & n. 87; see also Department of Justice & Federal Trade Commission, Horizontal Merger Guidelines § 4.1 (2010) (defining product markets and market power in terms of a firm’s ability to raise prices for consumers). Rather, broadband providers’ ability to impose restrictions on edge providers simply depends on end users, not being fully responsive to the imposition of such restrictions. See supra at 646. If the dissent believes that broadband providers’ ability to restrict edge-provider traffic without having their end users react would itself represent an exercise of market power, then the dissent’s dispute with the Commission’s reasoning appears to be largely semantic: the Commission expressly found that end users are not responsive in this fashion even if it never used the term “market power” in doing so. See Open Internet Order, 25 F.C.C.R. at 17924-25 ¶ 34.

Furthermore, the Commission established that the threat that broadband providers would utilize their gatekeeper ability to restrict edge-provider traffic is not, as the Commission put it, “merely theoretical.” Open Internet Order, 25 F.C.C.R. at 17925 ¶ 35. In support of its conclusion that broadband providers could and would act to limit Internet openness, the Commission pointed to four prior instances in which they had done just that. These involved a mobile broadband provider blocking online payment services after entering into a contract with a competing service; a mobile broadband provider restricting the availability of competing VoIP and streaming video services; a fixed broadband provider blocking VoIP applications; and, of course, Comcast’s impairment of peer-to-peer file sharing that was the subject of the Comcast Order. See id. Although some of these incidents may not have involved “adjudicated findings of misconduct,” as Verizon asserts, Verizon’s Br. 50, that hardly means that no record evidence supports the Commission’s conclusion that the incidents had in fact occurred. Likewise, the fact that we vacated the Comcast Order — rendering it, according to Verizon, a “legal nullity,” Verizon’s Br. 51 — did not require the Commission to entirely disregard the underlying conduct that produced that order. In Comcast, we held that the Commission had failed to cite any statutory authority that justified its order, not that Comcast had never impaired Internet traffic. See Comcast, 600 F.3d at 644. Nor, finally, did the Commission’s invocation of these examples demonstrate that it was attempting to “impose an ‘industry-wide solution for a problem that exists only in isolated pockets.’ ” Verizon’s Br. 51 (quoting Associated Gas Distributors v. FERC, 824 F.2d 981, 1019 (D.C.Cir.1987)). Rather, as the Commission explained, these incidents — which occurred “notwithstanding the Commission’s adoption of open Internet principles,” Commission enforcement proceedings against those who violated those principles, and specific Commission orders “requir[ing] certain broadband providers to adhere to open Internet obligations,” Open Internet Order, 25 F.C.C.R. at 17926-27 ¶ 37 — buttressed the agency’s conclusion that broadband providers’ incentives and ability to restrict Internet traffic could produce “[widespread interference with the Internet’s openness” in the absence of Commission action, id. at 17927 ¶ 38. Such a “problem” is doubtless “industry-wide.” Associated Gas Distributors, 824 F.2d at 1019.

Finally, Verizon argues that the Open Internet Order rules will necessarily have the opposite of their intended effect because they will “harm innovation and deter investment by increasing costs, foreclosing potential revenue streams, and restricting providers’ ability to meet consumers’ evolving needs.” Verizon’s Br. 52; see also Dissenting Op. at 666-67. In essence, Verizon believes that any stimulus to edge-provider innovation, as well as any consequent demand for broadband infrastructure, produced by the Open Internet Order will be outweighed by the diminished incentives for broadband infrastructure investment caused by the new limitations on business models broadband providers may employ to reap a return on their investment. As Verizon points out, two members of the Commission agreed that the rules would be counterproductive, and several commenters contended that certain regulations of broadband providers would run the risk of stifling infrastructure investment. See Open Internet. Order, 25 F.C.C.R. at 18054-56 (Dissenting Statement of Commissioner McDowell); id. at 18088-91 (Dissenting Statement of Commissioner Baker); Verizon Comments at 40-86, Docket No. 09-191 (Jan. 14, 2010); MetroPCS Comments at 24-35, Docket No. 09-191 (Jan 14, 2010); see also Open Internet Order, 25 F.C.C.R. at 17931 ¶ 42 n. 143 (discussing the comments of the Department of Justice and Federal Trade Commission).

The record, however, also contains much evidence supporting the Commission’s conclusion that, “[b]y comparison to the benefits of [its] prophylactic measures, the costs associated with the open Internet rules ... are likely small.” Open Internet Order, 25 F.C.C.R. at 17928 ¶ 39. This is, in other words, one of those cases — quite frequent in this circuit — where “the available data do[ ] not settle a regulatory issue and the agency must then exercise its judgment in moving from the facts and probabilities on the record to a policy conclusion.” State Farm, 463 U.S. at 52, 103 S.Ct. 2856. Here the Commission reached its “policy conclusion” by emphasizing, among other things, (1) the absence of evidence that similar restrictions of broadband providers had discouraged infrastructure investment, and (2) the strength of the effect on broadband investment that it anticipated from edge-provider innovation, which would benefit both from the preservation of the “virtuous circle of innovation” created by the Internet’s openness and the increased certainty in that openness engendered by the Commission’s rules. Open Internet Order, at 17928-31 ¶¶40-42. In so doing, the Commission has offered “a rational connection between the facts found and the choice made,” State Farm, 463 U.S. at 52, 103 S.Ct. 2856 (internal quotation marks omitted), and Verizon has given us no persuasive reason to question that judgment.

III.

Even though section 706 grants the Commission authority to promote broadband deployment by regulating how broadband providers treat edge providers, the Commission may not, as it recognizes, utilize that power in a manner that contravenes any specific prohibition contained in the Communications Act. See Open Internet Order, 25 F.C.C.R. at 17969 ¶ 119 (reiterating the Commission’s disavowal of “a reading of Section 706(a) that would allow the agency to trump specific mandates of the Communications Act”); see also D. Ginsberg & Sons, Inc. v. Popkin, 285 U.S. 204, 208, 52 S.Ct. 322, 76 L.Ed. 704 (1932) (“General language of a statutory provision, although broad enough to include it, will not be held to apply to a matter specifically dealt with in another part of the same enactment.”). According to Verizon, the Commission has done just that because the anti-discrimination and anti-blocking rules “subject[ ] broadband Internet access service ... to common carriage regulation, a result expressly prohibited by the Act.” Verizon’s Br. 14.

We think it obvious that the Commission would violate the Communications Act were it to regulate broadband providers as common carriers. Given the Commission’s still-binding decision to classify broadband providers not as providers of “telecommunications services” but instead as providers of “information services,” see supra at 630-31, such treatment would run afoul of section 153(51): “A telecommunications carrier shall be treated as a common carrier- under this [Act] only to the extent that it is engaged in providing telecommunications services.” 47 U.S.C. § 153(51); see also Wireless Broadband Order, 22 F.C.C.R. at 5919 ¶ 50 (concluding that a “service provider is to be treated as a common carrier for the telecommunications services it provides, but it cannot be treated as a common carrier with respect to other, non-telecommunications services it may offer, including information services”). Likewise, because the Commission has classified mobile broadband service as a “private” mobile service, and not a “commercial” mobile service, see Wireless Broadband Order, 22 F.C.C.R. at 5921 ¶ 56, treatment of mobile broadband providers as common carriers would violate section 332: “A person engaged in the provision of a service that is a private mobile service shall not, insofar as such person is so engaged, be treated as a common carrier for any purpose under this [Act].” 47 U.S.C. § 332(c)(2); see Cellco, 700 F.3d at 538 (“[M]obile-data providers are statutorily immune, perhaps twice over, from treatment as common carriers.”).

Insisting it has transgressed neither of these prohibitions, the Commission begins with the rather half-hearted argument that the Act referred to in sections 153(51) and 332 is the Communications Act of 1934, and that when the Commission utilizes the authority granted to it in section 706— enacted as part of the 1996 Telecommunications Act — it is not acting “under” the 1934 Act, and thus is “not subject to the statutory limitations on common-carrier treatment.” Commission’s Br. 68. But section 153(51) was also part of the 1996 Telecommunications Act. And regardless, “Congress expressly directed that the 1996 Act ... be inserted into the Communications Act of 1934.” AT & T Corp., 525 U.S. at 377, 119 S.Ct. 721 (citing Telecommunications Act of 1996 § 1(b)). The Commission cannot now so easily escape the statutory prohibitions on common carrier treatment.

Thus, we must determine whether the requirements imposed by the Open Internet Order subject broadband providers to common carrier treatment. If they do, then given the manner in which the Commission has chosen to classify broadband providers, the regulations cannot stand. We apply Chevron’s deferential standard of review to the interpretation and application of the statutory term “common carrier.” See Cellco, 700 F.3d at 544. After first discussing the history and use of that term, we turn to the issue of whether the Commission’s interpretation of “common carrier” — and its conclusion that the Open Internet Order’s rules do not constitute common carrier obligations — was reasonable.

A.

Offering little guidance as to the meaning of the term “common carriér,” the Communications Act defines that phrase, somewhat circularly, as “any person engaged as a common carrier for hire.” 47 U.S.C. § 153(11). Courts and the Commission have therefore resorted to the common law to come up with a satisfactory definition. See FCC v. Midwest Video Corp., 440 U.S. 689, 701 n. 10, 99 S.Ct. 1435, 59 L.Ed.2d 692 (1979) (“Midwest Video IP’).

In the Nineteenth Century, American courts began imposing certain obligations — conceptually derived from the traditional legal duties of innkeepers, ferrymen, and others who served the public— on companies in the transportation and communications industries. See Cellco, 700 F.3d at 545. As the Supreme Court explained in Interstate Commerce Commission v. Baltimore & Ohio Railroad Co., 145 U.S. 263, 275, 12 S.Ct. 844, 36 L.Ed. 699 (1892), “the principles of the common law applicable to common carriers ... demanded little more than that they should carry for all persons who applied, in the order in which the goods were delivered at the particular station, and that their charges for transportation should be reasonable.” Congress subsequently codified these duties, first in the 1887 Interstate Commerce Act, ch. 104, 24 Stat. 379, then the Manns-Elkins Act of 1910, ch. 309, 36 Stat. 539, and, most relevant , here, the Communications Act of 1934, ch. 652, 48 Stat. 1064. See Cellco, 700 F.3d at 545-46.

Although the nature and scope of the duties imposed on common carriers have evolved over the last century, see, e.g., Orloff v. FCC, 352 F.3d 415, 418-21 (D.C.Cir.2003) (discussing the implications of the relaxation of the tariff-filing requirement), the core of the common law concept of common carriage has remained intact. In National Association of Regulatory Utility Commissioners v. FCC, 525 F.2d 630, 642 (D.C.Cir.1976) (“NARUC I”), we identified the basic characteristic that distinguishes common carriers from “private” carriers — i.e., entities that are not common carriers — as “[t]he common law requirement of holding oneself out to serve the public indiscriminately.” “[A] carrier will not be a common carrier,” we further explained, “where its practice is to make individualized decisions, in particular cases, whether and on what terms to deal.” Id. at 641. Similarly, in National Association of Regulatory Utility Commissioners v. FCC, 533 F.2d 601, 608 (1976) (“NARUC II ”), we concluded that “the primary sine qua non of common carrier status is a quasi-public character, which arises out of the undertaking to carry for all people indifferently.” (Internal quotation marks omitted).

For our purposes, perhaps the seminal case applying this notion of common carriage is Midwest Video II. At issue in Midwest Video II was a set of regulations compelling cable television systems to operate a minimum number of channels and to hold certain channels open for specific users. 440 U.S. at 692-93, 99 S.Ct. 1435. Cable operators were barred from exercising any discretion over who could use those latter channels and what those users could transmit. They were also forbidden from charging users any fee for some of the channels and limited to charging an “appropriate” fee for the remaining channels. Id. at 693-94, 99 S.Ct. 1435. Because at that time the Commission had no express statutory authority over cable systems, it sought to justify these rules as ancillary to its authority to regulate broadcasting. Id. at 696-99, 99 S.Ct. 1435.

Rejecting this argument, the Supreme Court held that the Commission had no power to regulate cable operators in this fashion. The Court reasoned that if the Commission sought to exercise such ancillary jurisdiction over cable operators on the basis of its authority over broadcasters, it must also respect the specific statutory limits of that authority, as “without reference to the provisions of the Act directly governing broadcasting, the Commission’s jurisdiction ... would be unbounded.” Midwest Video II, 440 U.S. at 706, 99 S.Ct. 1435. Congress had expressly prohibited the Commission from regulating broadcasters as common carriers, a limitation that must then, according to the Court, also extend to cable operators. Id. at 707, 99 S.Ct. 1435. And the challenged regulations, the Court held, “plainly impose common-carrier obligations on cable operators.” Id. at 701, 99 S.Ct. 1435. In explaining this conclusion, the Court largely reiterated the nature of the obligations themselves: “Under the rules, cable systems are required to hold out dedicated channels on a first-come, nondiscriminatory basis. Operators are prohibited from determining or influencing the content of access programming. And the rules delimit what operators may charge for access and use of equipment.” Id. at 701-02, 99 S.Ct. 1435 (internal citations omitted).

In Célico, we recently confronted the similar question of whether a Commission regulation compelling mobile telephone companies to offer data roaming agreements to one another on “commercially reasonable” terms impermissibly regulated these providers as common carriers. 700 F.3d at 537. From the history and decisions surveyed above, we distilled “several basic principles” that guide our analysis here. Id. at 547. First, “[i]f a carrier is forced to offer service indiscriminately and on general terms, then that carrier is being relegated to common carrier status.” Id. We also clarified, however, that “there is an important distinction between the question whether a given regulatory regime is consistent with common carrier or private carrier status, and the Midwest Video II question whether that regime necessarily confers common carrier status.” Id. (internal citations omitted). Thus, “common carriage is not all or nothing — there is a gray area in which although a given regulation might be applied to common carriers, the obligations imposed are not common carriage per se.” Id. In this “space between per se common carriage and per se private carriage,” we continued, “the Commission’s determination that a regulation does or does not confer common carrier status warrants deference.” Id.

Given these principles, we concluded that the data roaming rule imposed no per se common carriage requirements because it left “substantial room for individualized bargaining and discrimination in terms.” Cellco, 700 F.3d at 548. The rule “expressly permitted] providers to adapt roaming agreements to ‘individualized circumstances without having to hold themselves out to serve all comers indiscriminately on the same or standardized terms.’ ” Id. That said, we cautioned that were the Commission to apply the “commercially reasonable” standard in a restrictive manner, essentially elevating it to the traditional common carrier “just and reasonable” standard, see 47 U.S.C. § 201(b), the rule might impose obligations that amounted to common carriage per se, a claim that could be brought in an “as applied” challenge. Cellco, 700 F.3d at 548^49.

B.

The Commission’s explanation in the Open Internet Order for why the regulations do not constitute common carrier obligations and its defense of -those regulations here largely rest on its belief that, with respect to edge .providers, broadband providers are not “carriers” at all. Stating that an entity is not a common carrier if it may decide on an individualized basis “ Vhether and on what terms to deal’ with potential customers,” the Commission asserted in the Order that “[t]he customers at issue here are the end users who subscribe to broadband Internet access services.” Open Internet Order, 25 F.C.C.R. at 17950-51 ¶ 79 (quoting NARUC I, 525 F.2d at 641) (emphasis added). It explained that because broadband providers would remain able to make “individualized decisions” in determining on what terms to deal with end users, the Order permitted the providers the “flexibility to customize service arrangements for a particular customer [that] is the hallmark of private carriage.” Id. at 17951 ¶ 79. Here, the Commission reiterates that “as long as [a broadband provider] is not required to serve end users indiscriminately, rules regarding blocking or charging edge providers do not create common carriage.” Commission’s Br. 61. We disagree.

It is true, generally speaking, that the “customers” of broadband providers are end users. But that hardly means that broadband providers could not also be carriers with respect to edge providers. “Since it is clearly possible for a given entity to carry on many types of activities, it is at least logical to conclude that one may be a common carrier with regard to some activities but not others.” NARUC II, 533 F.2d at 608. Because broadband providers furnish a service to edge providers, thus undoubtedly functioning as edge providers’ “carriers,” the obligations that the Commission imposes on broadband providers may well constitute common carriage per se regardless of whether edge providers are broadband providers’ principal customers. This is true whatever the nature of the preexisting commercial relationship between broadband providers and edge .providers. In contending otherwise, the Commission appears to misunderstand the nature of the inquiry in which we must engage. The question is not whether, absent the Open Internet Order, broadband providers would or did act as common carriers with respect to edge providers; rather, the question is whether, given the rules imposed by the Open Internet Order, broadband providers are now obligated to act as common carriers. See Midwest Video II, 440 U.S. at 701-02, 99 S.Ct. 1435.

In support of its understanding of common carriage, the Commission first invokes section 201(a), which provides that it is the “duty of every common carrier ... to furnish ... communication service upon reasonable request therefor.” 47 U.S.C. § 201(a). No one disputes that a broadband provider’s transmission of edge-provider traffic to its end-user subscribers represents a valuable service: an edge provider like Amazon wants and needs a broadband provider like Comcast to permit its subscribers to use Amazon.com. According to the Commission, however, because edge providers generally do not “request” service from broadband providers, and may have no direct relationship with end users’ local access providers, broadband providers cannot be common carriers with respect to such edge providers. But section 201(a) describes a “duty” of a common carrier, not a prerequisite for qualifying as a common carrier in the first place. ■ More important, the Open Internet Order imposes this very duty on broadband providers: given the Open Internet Order’s, anti-blocking and anti-discrimination requirements, if Amazon were now to make a request for service, Comcast must comply. That is, Comcast must now “furnish ... communication service '.upon reasonable request therefor.” 47 U.S.C. § 201(a).

Similarly flawed is the Commission’s argument that because the Communications Act defines a “common carrier” as a “common carrier for hire,” 47 U.S.C. § 153(11) (emphasis added), a common carrier relationship may exist only with respect to those customers who purchase service from the carrier. As Verizon aptly puts it in response, the fact that “broadband providers ... generally have not charged edge providers for access or offered them differentiated services ... has no legal significance because the avowed purpose of the rules is to deny providers the discretion to do so now and in the future.” Verizon’s Reply Br. 5 n. 3. In other words, but for the Open Internet Order, broadband providers could freely impose conditions on the nature and quality of the service they furnish edge providers, potentially turning certain edge providers — currently able to “hire” their service for free — into paying customers. The Commission may not claim that the Open Internet Order imposes no common carrier obligations simply because it compels an entity to continue furnishing service at no cost.

Likewise, the Commission misses the point when it contends that because the Communications Act “imposes non-discrimination requirements on many entities that are not common carriers,” the Order’s requirements cannot “transform[] providers into common carriers.” Commission’s Br. 66-67. In support, the Commission cites 47 U.S.C. § 315(b), which requires that broadcasters charge political candidates nondiscriminatory rates if broadcasters permit them to use their stations, as well as 47 U.S.C. § 548(c)(2)(B), which prohibits satellite programming vendors owned in part or in whole by a cable operator from discriminating against other cable operators in the delivery of programming. Commission’s Br. 66-67. But Congress has no statutory obligation to avoid imposing common carrier obligations on those who might not otherwise operate as common carriers, and thus the extent to which the cited provisions might regulate those entities as such is irrelevant. The Commission, on the other hand, has such an obligation with respect to entities it has classified as statutorily exempt from common carrier treatment, and the issue here is whether it has nonetheless “relegated [those entities], pro tanto, to common-carrier status.” Midwest Video II, 440 U.S. at 700-01, 99 S.Ct. 1435.

In these respects, Midwest Video II is indistinguishable. The Midwest Video II cable operators’ primary “customers” were their subscribers, who paid to have programming delivered to them in their homes. There, as here, the Commission’s regulations required the regulated entities to carry the content of third parties to these customers — content the entities otherwise could have blocked at their discretion. Moreover, much like the rules at issue here, the Midwest Video II regulations compelled the operators to hold open certain channels for use at no cost — thus permitting specified programmers to “hire” the cable operators’ services for free. Given that the cable operators in Midwest Video II were carriers with respect to these third-party programmers, we see no basis for concluding that broadband providers are not similarly carriers with respect to third-party edge providers.

The Commission advances several grounds for distinguishing Midwest Video II. None is convincing.

The Commission asserts that, unlike in Midwest Video II, here the content is delivered to end users only when an end user requests it — i.e., by clicking on a link to an edge provider’s website. But the same was essentially true in Midwest Video II: cable companies’ customers would not actually receive the content on the dedicated public access channels unless they chose to watch those channels. The access requested by the programmers in Midwest Video II, like the access requested by edge providers here, is the ability to have their communications transmitted to end-user subscribers if those subscribers so desire.

Nor, contrary to the Commission’s contention, is it at all relevant that in Midwest Video II only a limited number of cable channels were available, while in this case the number of edge providers a broadband provider could serve is unlimited. Whether an entity qualifies as a carrier does not turn on how much content it is able to carry or the extent to which other content might be crowded out. A short train is no more a carrier than a long train, or even a train long enough to serve every possible customer.

Finally, Midwest Video II cannot be distinguished on the basis that the Court there emphasized the degree to which the Commission’s rules impinged on cable operators’ “editorial discretion,” and “transferred control” over the content transmitted. Commission’s Br. 65. The Court made two related points regarding editorial discretion, neither of which helps the Commission. First, it observed that the need to protect editorial discretion was one reason Congress forbade common carrier treatment of broadcasters in the first place, a rationale that also applied to cable operators, thus confirming the Court’s decision to extend that statutory prohibition to the Commission’s attempt to exercise its ancillary jurisdiction over such entities. Midivest Video II, 440 U.S. at 700, 706-08, 99 S.Ct. 1435. Here, whatever might be the justifications for prohibiting common carrier treatment of “information service” providers and “commercial” mobile service providers, such treatment is undoubtedly prohibited. See 47 U.S.C. §§ 153(51), 332(c)(2). Second, the Court emphasized that, unlike the regulations approved in United States v. Midwest Video Corp., 406 U.S. 649, 92 S.Ct. 1860, 32 L.Ed.2d 390 (1972) (“Midwest Video I ”) — which required certain cable companies to create their own programming and maintain facilities for local production, id. at 653-55, 92 S.Ct. 1860 — the regulations in Midwest Video II “transferred control of the content of access cable channels from cable operators to members of the public.” Midwest Video II, 440 U.S. at 700, 99 S.Ct. 1435. The Court’s point was simply that the Midwest Video I regulations had created no common carrier obligations because they had imposed no obligation on cable operators to provide carriage to any third party. By giving third parties “control” over the transmissions that cable operators carried, however, the Midwest Video II regulations did. The regulations here accomplish the very same sort of transfer of control: whereas previously broadband providers could have blocked or discriminated against the content of certain edge providers, they must now carry the content those edge providers desire to transmit. The only remaining question, then, is whether the Open Internet Order’s rules have so limited broadband providers’ control over edge providers’ transmissions that the regulations constitute common carriage per se. It is to that question that we now turn.

C.

We have little hesitation in concluding that the anti-discrimination obligation imposed on fixed broadband providers has “relegated [those providers], pro tanto, to common carrier status.” Midwest Video II, 440 U.S. at 700-01, 99 S.Ct. 1435. In requiring broadband providers to serve all edge providers without “unreasonable discrimination,” this rule by its very terms compels those providers to hold themselves out “to serve the public indiscriminately.” NARUC I, 525 F.2d at 642.

Having relied almost entirely on the flawed argument that broadband providers are not carriers with respect to edge providers, the Commission offers little response on this point. In its briefs, the Commission contends only that if the Open Internet Order imposes common carriage requirements, so too would the regulations at issue in United States v. Southwestern Cable Co., 392 U.S. 157, 88 S.Ct. 1994, 20 L.Ed.2d 1001 (1968), which the Supreme Court declined to strike down. Southwestern Cable involved a Commission rule that, among other things, compelled cable operators to transmit the signals of local broadcasters when cable operators imported the competing signals of other broadcasters into the local service area. Id. at 161, 88 S.Ct. 1994. Such a rule is plainly distinguishable from the Open Internet Order’s anti-discrimination rule because the Southwestern Cable regulation imposed no obligation on cable operators to hold their facilities open to the public generally, but only to certain specific broadcasters if and when the cable operators acted in ways that might harm those broadcasters. As the Court later explained in Midwest Video II, the Southwestern Cable rule “was limited to remedying a specific perceived evil,” and “did not amount to a duty to hold out facilities indifferently for public use.” 440 U.S. at 706 n. 16, 99 S.Ct. 1435. The Open Internet Order’s anti-discrimination provision is not so limited, as the compelled carriage obligation applies in all circumstances and with respect to all edge providers.

Significantly for our purposes, the Commission never argues that the Open Internet Order’s “no unreasonable discrimination” standard somehow differs from the nondiscrimination standard applied to common carriers generally — the argument that salvaged the data roaming requirements in Célico. In a footnote in the Order itself, the Commission suggested that it viewed the rule’s allowance for “reasonable network management” as establishing treatment that was somehow inconsistent with per se common carriage. See Open Internet Order, 25 F.C.C.R. at 17951 ¶ 79 n. 251. But the Commission has forfeited this argument by failing to raise it in its briefs here. See Comcast, 600 F.3d at 660; Roth v. U.S. DOJ, 642 F.3d 1161, 1181 (D.C.Cir.2011).

In any event, the argument is without merit. The Order defines the “reasonable network management” concept as follows: “A network management practice is reasonable if it is appropriate and tailored to achieving a legitimate network management purpose, taking into account the particular network architecture and technology of the broadband Internet access service.” Open Internet Order, 25 F.C.C.R. at 17952 ¶ 82. This provision, the Commission explained, would permit broadband providers to do two things, neither of which conflict with per se common carriage. First, “the reasonable network management” exception would permit broadband providers to “address[] traffic that is unwanted by end users ... such as by providing services or capabilities consistent with an end user’s choices regarding parental controls or security capabilities.” Id. Because the relevant service broadband providers furnish to edge providers is the ability to access end users if those end users so desire, a limited exception permitting end users to direct broadband providers to block certain traffic by no means detracts from the common carrier nature of the obligations imposed on broadband providers. Second, the Order defines “reasonable network management” to include practices designed to protect the network itself by “addressing traffic that is harmful to the network” and “reducing or mitigating the effects of congestion.” Id. at 17952 ¶ 82. As Verizon correctly points out, however, this allowance “merely preserves a common carrier’s traditional right to ‘turn[] away [business] either because it is not of the type normally accepted or because the carrier’s capacity has been exhausted.’ ” Verizon’s Br. 20 (quoting NARUC I, 525 F.2d at 641). Railroads have no obligation to allow passengers to carry bombs on board, nor need they permit passengers to stand in the aisles if all seats are taken. It is for this reason that the Communications Act bars common carriers from engaging in “unjust or unreasonable discrimination,” not all discrimination. 47 U.S.C. § 202 (emphasis added).

The Commission has provided no basis for concluding that in permitting “reasonable” network management, and in prohibiting merely “unreasonable” discrimination, the Order’s standard of “reasonableness” might be more permissive than the quintessential common carrier standard. See Cellco, 700 F.3d at 548 (characterizing the “just and reasonable” standard as being that “applicable to common carriers”). To the extent any ambiguity exists regarding how the Commission will apply these rules in practice, we think it is best characterized as ambiguity as to how the common carrier reasonableness standard applies in this context, not whether the standard applied is actually the same as the common carrier standard. Unlike the data roaming requirement at issue in Célico, which set forth a “commercially reasonable” ' standard, see id. at 537, the language of the Open Internet Order’s anti-discrimination rule mirrors, almost precisely, section 202’s language establishing the basic common carrier obligation not to “make any unjust or unreasonable discrimination.” 47 U.S.C. § 202. Indeed, confirming that the two standards are equivalent, the Commission responded to commenters who argued that the “no unreasonable discrimination” requirement was too vague by quoting another commenter who observed that “[s]eventy-five years of experience have shown [the ‘unreasonable’ qualifier in Section 202] to be both administrable and indispensable to the sound administration of the nation’s telecommunications laws.” Open Internet Order, 25 F.C.C.R. at 17949 ¶77 n. 240. Moreover, unlike the data roaming rule in Cellco — which spelled out “sixteen different factors plus a catchall ... that the Commission must take into account in evaluating whether a proffered roaming agreement is commercially reasonable,” thus building into the standard “considerable flexibility,” Cellco, 700 F.3d at 548 — the Open Internet Order makes no attempt to ensure that its reasonableness standard remains flexible. Instead, with respect to broadband providers’ potential negotiations with edge providers, the Order ominously declares: “it is unlikely that pay for priority would satisfy the ‘no unreasonable discrimination’ standard.” Open Internet Order, 25 F.C.C.R. at 17947 ¶ 76. If the Commission will likely bar broadband providers from charging edge providers for using their service, thus forcing them to sell this service to all who ask at a price of $0, we see no room at all for “individualized bargaining.” Célico, 700 F.3d at 548.

Whether the Open Internet Order ’s anti-blocking rules, applicable to both fixed and mobile broadband providers, likewise establish per se common carrier obligations is somewhat less clear. According to Verizon, they do because they deny “broadband providers discretion in deciding which traffic from ... edge providers to carry,” and deny them “discretion over carriage terms by setting a uniform price of zero.” Verizon’s Br. 16-17. This argument has some appeal. The anti-blocking rules establish a minimum level of service that broadband providers must furnish to all edge providers: edge providers’ “content, applications [and] services” must be “effectively [ ]usable.” Open Internet Order, 25 F.C.C.R. at 17943 ¶ 66. The Order also expressly prohibits broadband providers from charging edge providers any fees for this minimum level of service. Id. at 17943-44 ¶ 67. In requiring that all edge providers receive this minimum level of access for free, these rules would appear on their face to impose per se common carrier obligations with respect to that minimum level of service. See Midwest Video II, 440 U.S. at 701 n. 9, 99 S.Ct. 1435 (a carrier may “operate as a common carrier with respect to a portion of its service only”).

At oral argument, however, Commission counsel asserted that “[i]t’s not common carriage to simply have a basic level of required service if you can negotiate different levels with different people.” Oral Arg. Tr. 86. This contention rests on the fact that under the anti-blocking rules broadband providers have no obligation to actually provide any edge provider with the minimum service necessary to satisfy the rules. If, for example, all edge providers’ “content, applications [and] services” are “effectively usable,” Open Internet Order, 25 F.C.C.R. at 17943 ¶ 66, at download speeds of, say, three mbps, a broadband provider like Verizon could deliver all edge providers’ traffic at speeds of at least four mbps. Viewed this way, the relevant “carriage” broadband providers furnish might be access to end users more generally, not the minimum required service. In delivering this service, so defined, the anti-blocking rules would permit broadband providers to distinguish somewhat among edge providers, just as Commission counsel contended at oral argument. For example, Verizon might, consistent with the anti-blocking rule — and again, absent the anti-discrimination rule — charge an edge provider like Netflix for high-speed, priority access while limiting all other edge providers to a more standard service. In theory, moreover, not only could Verizon negotiate separate agreements with each individual edge provider regarding the level of service provided, but it could also charge similarly-situated edge providers completely different prices for the same service. Thus, if the relevant service that broadband providers furnish is access to their subscribers generally, as opposed to access to their subscribers at the specific minimum speed necessary to satisfy the anti-blocking rules, then these rules, while perhaps establishing a lower limit on the forms that broadband providers’ arrangements with edge providers could take, might nonetheless leave sufficient “room for individualized bargaining and discrimination in terms” so as not to run afoul of the statutory prohibitions on common carrier treatment. Célico, 700 F.3d at 548.

Whatever the merits of this view, the Commission advanced nothing like it either in the underlying Order or in its briefs before this court. Instead, it makes no distinction at all between the anti-discrimination and anti-blocking rules, seeking to justify both types of rules with explanations that, as we have explained, are patently insufficient. We are unable to sustain the Commission’s action on a ground upon which the agency itself never relied. Lacson v. Department of Homeland Security, 726 F.3d 170, 177 (D.C.Cir. 2013); see also United States v. Southerland, 486 F.3d 1355, 1360 (D.C.Cir.2007) (“argument[s] • • • raised for the first time at oral argument [are] forfeited”). Nor may we defer to a reading of a statutory term that the Commission never offered. Shieldalloy Metallurgical Corp. v. Nuclear Regulatory Comm’n, 624 F.3d 489, 495 (D.C.Cir.2010).

The disclosure rules are another matter. Verizon does not contend that these rules, on their own, constitute per se common carrier obligations, nor do we see any way in which they would. Also, because Verizon does not direct its First Amendment or Takings Clause claims against the disclosure obligations, we have no need to address those contentions here.

Verizon does argue that the disclosure rules are not severable, insisting that if the anti-discrimination and anti-blocking rules fall so too must the disclosure requirements. We disagree. ‘Whether the offending portion of a regulation is severable depends upon the intent of the agency and upon whether the remainder of the regulation could function sensibly without the stricken provision,” MD/DC/DE Broadcasters Ass’n v. FCC, 236 F.3d 13, 22 (D.C.Cir.2001) (emphasis omitted). At oral argument, Commission counsel explained that the rules function separately, Oral Arg. Tr. 81-82, and we are satisfied that the Commission would have adopted the disclosure rules absent the rules we now vacate, which, we agree, operate independently. See Davis County Solid Waste Management v. EPA, 108 F.3d 1454, 1457-59 (D.C.Cir.1997) (finding promulgated standard to be severable where EPA asserted in rehearing petition that, contrary to its position at oral argument, the standards could stand alone).

IV.

For the forgoing reasons, although we reject Verizon’s challenge to the Open Internet Order’s disclosure rules, we vacate both the anti-discrimination and the anti-blocking rules. See Northern Air Cargo v. U.S. Postal Service, 674 F.3d 852, 860-61 (D.C.Cir.2012) (appropriateness of vacatur dependent on whether “(1) the agency’s decision is so deficient as to raise serious doubts whether the agency can adequately justify its decision at all; and (2) vacatur would be seriously disruptive or costly”); Comcast, 600 F.3d at 661 (vacating the Comcast Order). We remand the case to the Commission for further proceedings consistent with this opinion.

So ordered.

SILBERMAN, Senior Circuit Judge,

concurring in part and dissenting in part:

I am in general agreement with the majority’s conclusion that the Open Internet Order impermissibly subjects broadband providers to treatment as common carriers, but I disagree with the majority’s conclusion that § 706 otherwise provides the FCC with affirmative statutory authority to promulgate these rules. I also think the Commission’s reasoning violates the Administrative Procedure Act. These differences are important since the majority opinion suggests possible regulatory modifications that might circumvent the prohibition against common carrier treatment.

I.

The Commission’s net neutrality regulation is purportedly designed to promote innovation among edge providers who, in turn, provide Internet user experience, thereby increasing user demand for broadband service and, ultimately, encouraging broadband providers to invest in infrastructure development to meet that demand. Open Internet Order, 25 F.C.C.R. 17905, 17907 ¶ 13 (2010). Verizon describe this theory as a “triple cushion shot.” As I will show, whatever its logic, it is based on a faulty factual premise. But my first disagreement with the Commission, and the majority, is to the claimed statutory authority.

I quite agree with the majority that the relevant statutory language is § 706 of the Communications Act. 47 U.S.C. § 1302. Although the FCC purports to rely on a scatter shot of other provisions of the statute, as well as § 706, none of those other provisions truly bear on the issue. “Emanations from the penumbra” may once have served to justify constitutional interpretation, but it hasn’t caught on as legitimate statutory interpretation. I also agree with the majority — and disagree with Verizon — that § 706 is a grant of positive regulatory authority, but it doesn’t come close to sanctioning the Commission’s regulation.

The statute directs the Commission to “encourage the deployment on a reasonable and timely basis of advanced telecommunications capability to all Americans ... by utilizing ... price cap regulation, regulatory forbearance, measures that promote competition in the local telecommunications market, or other regulating methods that remove barriers to infrastructure investment.” 47 U.S.C. § 1302(a).1

The FCC contends for, and the majority grants, Chevron deference as to the interpretation of this language. I don’t disagree that Chevron is called for, but Chevron “is not a wand by which courts can turn an unlawful frog into a legitimate prince.” Associated Gas Distributors v. F.E.R.C., 824 F.2d 981, 1001 (D.C.Cir.1987).

The key words obviously are “measures that promote competition in the local telecommunications market or other regulating methods that remove barriers to infrastructure investment.” Those are the words that grant actual authority. Yet the Commission does not ground its regulation on this language. Indeed, both the Commission and the majority conflate these two clauses, though they have distinct functions. “Promoting competition in the telecommunications market” implies a regulation that encourages broadband providers to compete with each other, head-to-head, on price and quality. Removing “barriers to infrastructure investment,” on the other hand, does not necessarily require any increased competition in the telecommunications market.2 For example, if a particular broadband provider were a monopolist, then by regulating its prices, the Commission might encourage it to expand supply to increase profits, rather than artificially restrict supply so as to charge supracompetitive rates. Such a regulation would not increase competition, but it would at least potentially remove a barrier to investment. This is, essentially, the theory that the Commission purportedly relies on: If the Commission theoretically could spur demand for broadband, the Commission would encourage further infrastructure investment regardless of head-to-head competition. Thus, it is on the “removing barriers” clause, primarily,3 that the Order must stand or fall. Yet, the Commission never actually identifies any practices of the broadband providers as “barriers to investment” — not once in over 100 pages — probably because it would be so far fetched an interpretation of those words.

Nor does the Commission state (or argue in its brief), contrary to the majority’s opinion, that the “triple cushion shot” — the means by which the Commission hopes to stimulate demand for better broadband— is designed to increase competition in the broadband market. See Majority Op. at 642 — 43 (citing 25 F.C.C.R. at 17910-11, 17970 ¶¶ 14, 120). Paragraph 14 makes no reference to competition,4 and paragraph 120 does not refer to competition between broadband providers in the local telecommunications market — which is the statutory objective. Indeed, paragraph 120 indicates that the Commission’s objective is to protect the edge providers (not in the telecommunications market) from content competition with the broadband providers.5

Indeed, the Commission frankly admits its purpose is much wider than the statutory objectives. It claims it must regulate broadly, so as to “proteet[ ] consumer choice, free expression, end-user control, and the ability to innovate without permission,” 25 F.C.C.R. at 17949 ¶78, which certainly indicates a Commission objective that exceeds the statutory authority granted in § 706.

The majority takes the statutory language even further; it states that the Commission’s

authority to promulgate regulations that promote broadband deployment encompasses the power to regulate broadband providers’ economic relationships with edge providers if, in fact, the nature of those relationships influences the rate and extent to which broadband providers develop and expand services for end users.

Majority Op. at 643. So much for the terms “promote competition in the local telecommunications market” or “remove barriers to infrastructure investment.” Presto, we have a new statute granting the FCC virtually unlimited power to regulate the Internet. This reading of § 706, as we said in Comcast Corp. v. FCC, “would virtually free the Commission from its congressional tether.” 600 F.3d 642, 655 (D.C.Cir.2010). The limiting principles the majority relies on are illusory.

The majority claims that the Commission cannot exceed its subject-matter jurisdiction over “interstate and foreign communication by wire and radio.” 25 F.C.C.R. at 17970 ¶ 121 (citing 47 U.S.C. § 152(a)). This is obviously true, but it is not a limitation on the Commission’s interpretation of this specific statutory provision. The question is not whether the statute permits the Commission to do absolutely anything — of course it does not— but, rather, whether § 706 contains any intrinsic limitations. If the Commission’s subject matter jurisdiction is a “limiting principle,” then we might as well call the First Amendment a limiting principle, for surely the Commission could not censor the Internet, even if doing so did somehow increase broadband deployment.

According to the majority, the Commission is also restrained because it may only regulate pursuant to § 706 if it does so to achieve a particular purpose: to “encourage the deployment on a reasonable and timely basis of advanced telecommunications capability to all Americans.” 25 F.C.C.R. at 17970 ¶ 121 (citing 47 U.S.C. § 1302(a)). This is an almost meaningless limitation, as demonstrated by the Open Internet Order itself. The Commission’s theory is that an open Internet will spur demand for broadband infrastructure. Id. at 17907 ¶ 3. But any regulation that, in the FCC’s judgment might arguably make the Internet “better,” could increase demand. I do not see how this “limitation” prevents § 706 from being carte blanche to issue any regulation that the Commission might believe to be in the public interest.

To sum up, § 706 requires the Commission to identify a “barrier[ ] to infrastructure investment” or a measure that “promoted competition” in the broadband market — which it has not.

II.

Verizon alternatively argue that, even assuming that § 706 grants the Commission its claimed authority, the regulation is arbitrary and capricious because its findings — such as they are — lack substantial evidence. I agree. Although we are not faced with a formal adjudication which would be judged by substantial evidence on a closed record, factual determinations that underly regulations must still be premised on demonstrated — and reasonable — evidential support. See Motor Vehicle Mfrs. Ass’n of U.S., Inc. v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43, 103 S.Ct. 2856, 77 L.Ed.2d 443 (1983).

The Commission purports to fear that broadband providers might discriminate against, or even block, the Internet traffic of specific edge providers or classes of edge providers, perhaps because broadband providers offer some competing services or because they might charge certain edge providers for premium services. The majority puts it even more starkly, asserting that the Commission “found that broadband providers have the technical and economic ability to impose restrictions on edge providers.” Majority Op. at 646 (emphasis added). But the Commission never actually made such a finding. Its conclusions are littered with “may,” “if,” and “might.” For example, according to the Commission, a broadband provider:

• “may have economic incentives to block or otherwise disadvantage specific edge providers”

• “might use this power to benefit its own or affiliated offerings at the expense of unaffiliated offerings”

• “may act to benefit edge providers that have paid it to exclude rivals”

• “may have incentives to increase revenues by charging edge providers”6

• “might withhold or decline to expand capacity in order to ‘squeeze’ non-prioritized traffic”

25 F.C.C.R. at 17915-22 ¶¶ 21-29. To be sure, the majority correctly observes that we should defer to an agency’s “predictive judgments as to the economic effect of a rule,” National Telephone Cooperative Ass’n v. FCC, 563 F.3d 536, 541 (D.C.Cir.2009), but deference to such a judgment must be based on some logic and evidence, not sheer speculation. That a party “may” do something is hardly a finding — at least in American law — that a party has done or will do something. Moreover, whether or not the “triple cushion shot” theory is rational economics (and I have my doubts), it rests, as I have noted, on a false factual premise — that the evidence supports a finding that broadband providers across the board, in all markets, enjoy sufficient, economic clout to take the above actions.

The Commission asserts — and the majority accepts — that broadband providers act as “gatekeepers” because each one has a so-called “terminating monopoly” over access to particular end users. These are terms, largely invented,7 the economic significance of which the Commission does not explain. All retail stores, for instance, are “gatekeepers.” The term is thus meaningful only insofar as the gatekeeper by means of a powerful economic position vis-a-vis consumers gains leverage over suppliers.8 The Commission made no effort to construct an analytic framework to measure this supposed gateway advantage — it is a rather slippery concept — nor did it adduce evidence to establish the economic power it would supposedly afford all broadband providers against all edge providers.

Without broadband provider market power, consumers, of course, have options; they can go to another broadband provider if they want to reach particular edge providers or if their connections to particular edge providers have been degraded. The Commission implicitly recognizes this, because it justifies exempting dial-up Internet providers from the Order by noting that “telephone service has historically provided the easy ability to switch among competing dial-up Internet access services.” 25 F.C.C.R. at 17935 ¶51. The Commission also exempts “backbone” Internet providers — which interconnect between broadband providers — obviously for the same reason. On the other hand, the Commission asserts that broadband customers may have few alternatives or they may be locked into long-term contracts with early-termination fees. To be sure, some difficulty switching broadband providers is certainly a factor that might contribute to a firm’s having market power, but that itself is not market power. There are many industries in which switching between competitors is not instantly achieved, but those industries may still be heavily disciplined by competitive forces because consumers will switch unless there are real barriers. By pointing to potential difficulties consumers may encounter switching broadband providers, the Commission is simply implying that broadband providers have market power (market power lite?), without actually examining if and where they do.

Although Verizon was reluctant to concede that even if a broadband provider had market power that would authorize the Commission to take action under § 706— presumably because they challenged any regulatory authority under § 706 — they did bring to our attention a Justice Department submission, discussed infra, that emphasized the necessity of the Commission limiting its regulatory initiatives to the control of broadband market power. Ex Parte Submission of the U.S. DOJ at 28, Docket No. 09-51 (Jan. 4, 2010). My discussion of market power reflects my view (and apparently the Justice Department’s) of what evidence would be adequate to support the Commission’s rule. In any event, Verizon certainly challenged the factual basis of the Commission’s “gateway” conclusion, so I don’t think the existence vel non of market power is really a different consideration. See Majority Op. at 647-48.

The majority does contend that four possible instances of broadband providers restricting users’ access to certain edge providers are sufficient evidence of broadband providers’ “incentives and ability to restrict Internet traffic.” Majority Op. at 649. That the Commission was able to locate only four potential examples of such conduct is, frankly, astonishing. In such a large industry where, as Verizon notes, billions of connections are formed between users and edge providers each year, one would think there should be ample examples of just about any type of conduct. But even if examples of such conduct were more numerous, it would still not be evidence that broadband providers are economically capable of restricting consumer choice. And, as the Commission noted, there are potentially efficient, pro-consumer reasons that an individual broadband provider might wish to restrict access to some edge providers. See 25 F.C.C.R. at 17921 ¶28 n. 80 (“Economics literature recognizes that access charges could be harmful under some circumstances and beneficial under others.... [T]he economic literature on two-sided markets is at an early stage of development.”). The Commission’s anecdotes then do not show that any broadband providers are capable of actually causing the harm about which the Commission is concerned.

My view, then, is that the Commission’s failure to conduct a market power analysis is fatal to its attempt to regulate, because it means that there is inadequate evidence to support the lynchpin of the Commission’s economic theory. The Commission actually recognized that a finding of market power would enhance its theory. 25 F.C.C.R. at 17923 ¶ 32. Indeed! But such a finding would, of course, have to be made market to market (indeed the statute specifically references local telecommunications markets), and if so, it would be a finding of a barrier to broadband investment without the mental gymnastics of the triple cushion shot. If one (or two) broadband providers have market power in any particular market and thereby could raise prices while restricting supply, the Commission could well conclude that was a barrier to broadband investment.

: Of course, before the Commission could determine whether a particular broadband provider possesses market power, it would have to first define the relevant market. Instead, the Commission, in this case, simply cited a 2009 study that found that “nearly 70 percent of households lived in census tracts where only one or two wire-line or fixed wireless firms provided advertised download speeds of at least 3 Mbps and upload speeds of at least 768 Kbps.” 25 F.C.C.R. at 17923 ¶ 32. Why are these speeds relevant? Because the Commission has previously, as part of its statutory duty to assess the state of broadband deployment, defined “broadband” to mean download speeds of at least 4 Mbps and upload speeds of at least 1 Mbps. Sixth Broadband Deployment, Report, 25 F.C.C.R. 9556, 9559 ¶ 5 (2010). According to the Commission, it is the minimum speed necessary to stream high quality video while simultaneously browsing the Internet and using email. Id. I don’t dispute the legitimacy of that definition. Yet, while the Commission is free to rely on technical considerations in defining the statutory term “broadband,” such considerations are irrelevant when it comes to defining the market in economic terms. A broadband provider offering a 2 Mbps connection is not, according to the FCC, really offering broadband. But it is quite likely that consumers, in deciding which Internet service to purchase, will compare products at varying speeds and price points. Slower service providers can still exert competitive pressure on faster service providers. So, too, can mobile broadband providers. Before the Commission can conclude that a market is concentrated, it must first define that market. It has made no effort to do so.

The Commission, moreover, does not address whether the trend in the broadband market is towards more or less competítion. Obviously the deployment of broadband infrastructure is a capital-intensive process, and it should not be surprising if, during a period of expansion, some areas are served by fewer competitors than others. But there is no evidence in the record suggesting that broadband providers are carving up territory or avoiding head-to-head competition. At least anecdotally, the opposite seems to be true. Google has now entered the broadband market as a direct competitor:

Google’s ultra-high-speed Internet service may finally be scaring the big Internet providers into action. Following Google’s announcement that it will expand into Austin, Texas, AT & T announced it will offer fiber Internet in the city, and Time Warner Cable announced it would offer citywide wireless Internet service.

But smaller companies are also trying to head off Google before the company even makes an announcement in their communities. This week, for example, the Lawrence, Kansas-based Internet provider Wicked Broadband began taking pre-orders for a residential fiber Internet service with speeds to rival Google Fiber’s.

Klint Finley, Google Fiber Spurs Monu- and-Pop Net Providers Too, Wired, Apr. 26, 2013, http://www.wired.com/wired enterprise/2013/04/google-fiberwicked/.

The Commission apparently wanted to avoid a disciplined inquiry focused on market power, notwithstanding the warning it received from the Justice Department less than a year before the regulation issued— which, as I noted, Verizon cited — a warning that unless the FCC’s focus was on market power, any regulation could actually discourage broadband development, thus frustrating the statutory objective:

Although enacting some form of regulation to prevent certain providers from exercising monopoly power may be tempting with regard to ... areas [served by only one or two broadband providers], care must be taken to avoid stifling the infrastructure investments needed to expand broadband access. In particular, price regulation would be appropriate only where necessary to protect consumers from the exercise of monopoly power and where such regulation would not stifle incentives to invest in infrastructure deployment.

Ex Parte Submission of the. U.S. DOJ at 28, Docket No. 09-51 (Jan. 4, 2010).

The Commission did postulate one other economic theory supposedly establishing a “barrier to infrastructure investment” that does not depend on the broadband providers possessing market power. It argued, essentially, that innovation among edge providers is a public good in that every broadband provider benefits from an open Internet, but each broadband provider has an individual incentive to charge edge providers for service because, if broadband providers were to forego that revenue stream, they would be unable to internalize all of the supposed benefits to innovation. 25 F.C.C.R. at 17919 ¶25. In short, the Commission speculates that the Open Internet Order prevents a classic “tragedy of the commons” — a situation in which each economic actor, behaving in his own self-interest, contributes to the destruction of a public good. See Garrett Hardin, The Tragedy of the Commons, 162 Science 1243 (1968). In such a situation, each actor would be better off if a central regulator prevented them from doing what would be in their private interest if they were acting unilaterally. Again, however, the Commission fails to make any real economic findings regarding whether these rules are actually necessary to prevent such a situation. As such, it is the sheerest of fanciful speculation.

Indeed, if a tragedy of the commons were likely in the broadband market, then one would expect Verizon and other broadband providers to support the Open Internet Order, because such a situation would be economically harmful to them in the long run. By the same token, when firms oppose, on antitrust grounds, the merger of competing firms, it is generally a reliable indicator that the merger is pro-competitive. See Frank H. Easterbrook, The Limits of Antitrust, 63 Tex. L.Rev. 1, 18 (1984) (“When a business rival brings suit, it is often safe to infer that the arrangement is beneficial to consumers.”). Firms can generally be relied upon to know their own best interest.

Perhaps most troubling, the Commission fails to appreciate the long-term impact of its own regulations. An unwarranted government interference in a functioning market is likely to persist indefinitely, whereas a failure to intervene, even when regulation would be helpful, is likely to be only temporarily harmful because new innovations are constantly undermining entrenched industrial powers. See id. at 3 (“[Jjudicial errors that tolerate baleful practices are self-correcting while erroneous condemnations are not.”); Tim Wu, The Master Switch 11 (2010) (“But as we have said, that which is centralized also eventually becomes a target for assault[.]”).

Nevertheless, the Commission justifies its aggressive, prophylactic regulation by asserting that the negative consequences of regulation (preserving the status quo) are likely to be minor, while the consequences of allowing the broadband market to evolve without regulation could be drastic and permanent. 25 F.C.C.R. at 17909 ¶ 12. I think this is quite wrong, but in any event, the agency’s judgment about the propriety of leaping before looking cannot displace the judgment of Congress which, in enacting § 706, did not so broadly empower the Commission. Rather, Congress required the agency to identify an actual barrier to infrastructure investment or a threat to competition, and the agency must have evidence that the barrier or threat exists.

III.

Because the Open Internet Order obviously imposes common carrier obligations on broadband providers, I join generally the opinion of the Court with respect to Part III. Indeed, even noted proponents of “net neutrality” acknowledge as much: “[N]et neutrality is the twenty-first century’s version of common carriage.... In the case of the Internet, common carriage under the name of net neutrality amounts to an FCC rule that bans any degree of blocking individual sites, [or] transmission of data.” Tim Wu, The Master Switch 236 (2010).

I have, however, one quibble with the majority’s analysis of the anti-blocking rules. Although ultimately concluding that the anti-blocking rules are unlawful, the majority says that whether those rules “likewise establish per se common carrier obligations is somewhat less clear.” Majority Op. at 657. Although the Order states that, under the anti-blocking rules, broadband providers may not degrade content so as to make it “effectively unusable,” the majority supposes that a broadband provider might voluntarily choose to offer service that is faster than the anti-blocking rules require, i.e., faster than the minimum speed necessary to make each edge provider effectively usable by consumers. By exceeding the minimum level of service, the majority suggests, the broadband providers would have wide latitude to engage in individualized bargaining, which might take this rule outside of common carriage per se. My concern with this hypothesis is that the phrase “effectively unusable” is subject to manipulation. I think it should mean that whatever speed is generally offered to most edge providers is the minimum necessary to be effectively usable. After all, it is artificial to distinguish between what is “effective” and what consumers expect. If a faster speed were to become standard, we would likely consider a slower speed to be effectively unusable. Thus, while there is a possibility that a “fast lane” Internet service might be offered on a non-common carriage, basis, the service that most users receive under this rule would still have to be offered as common carriage, at a regulated price of zero. In any event, as the majority recognizes, the Commission did not make this argument, so the anti-blocking rules must fall.9

This regulation essentially provides an economic preference to a politically powerful constituency, a constituency that, as is true of typical rent seekers, wishes protection against market forces. The Commission does not have authority to grant such a favor.

1

. Because § 706(b) contains almost the same language, it is unnecessary to discuss these two provisions separately. See 47 U.S.C. § 1302(b) (The Commission "shall take immediate action ... by removing barriers to infrastructure investment and by promoting competition in the telecommunications market.”).

2

. An example of a paradigmatic barrier to infrastructure investment would be state laws that prohibit municipalities from creating their own broadband infrastructure to compete against private companies. See Klint Finley, Why Your City Should Compete With Google’s Super-Speed Internet, Wired, May 28, 2013, http://www.wired.com/wiredenterprise/ 2013/05/community-fiber/.

3

.The transparency rules at least have the added benefit of facilitating consumer choice by providing information, which could lead to greater competition in the broadband market.

4

. The Internet’s openness is critical to these outcomes, because it enables a virtuous circle of innovation in which new uses of the network — including new content, applications, services, and devices — lead to increased end-user demand for broadband, which drives network improvements, which in turn lead to further innovative network uses. Novel, improved, or lower-cost offerings introduced by content, application, service, and device providers spur end-user demand and encourage broadband providers to expand their networks and invest in new broadband technologies. Streaming video and e-commerce applications, for instance, have led to major network improvements such as fiber to the premises, VDSL, and DOCSIS 3.0. These network improvements generate new opportunities for edge providers, spurring them to innovate further.. Each round of innovation increases the value of the Internet for broadband providers, edge providers, online businesses, and consumers. Continued operation of this virtuous circle, however, depends upon low barriers to innovation and entry by edge providers, which drive end-user demand. Restricting edge providers’ ability to reach end users, and limiting end users’ ability to choose which edge providers to patronize, would reduce the rate of innovation at the edge and, in turn, the likely rate of improvements to network infrastructure. Similarly, restricting the ability of broadband providers to put the network to innovative uses may reduce the rate of improvements to network infrastructure.

25- F.C.C.R. at 17910-11 ¶ 14.

5

. In directing the Commission to "encourage the deployment on a reasonable and timely basis of advanced telecommunications capability to all Americans ... by utilizing ... price cap regulation, regulatory forbearance, measures that promote competition in the local telecommunications market, or other regulating methods that remove barriers to infrastructure investment," Congress necessarily invested the Commission with the statutory authority to carry out those acts. Indeed, the relevant Senate Report explained that the ’ provisions of Section 706 are "intended to ensure that one of the primary objectives of the [1996 Act] — to accelerate deployment of advanced telecommunications capability — is achieved,” and stressed that these provisions are “a necessary fail-safe” to guarantee that Congress’s objective is reached. It would be odd indeed to characterize Section 706(a) as a "fail-safe” that "ensures” the Commission’s ability to promote advanced services if it conferred no actual authority. Here, under our reading, Section 706(a) authorizes the Commission to address practices, such as blocking VoIP communications, degrading or raising the cost of online video, or denying end users material information about their broadband service, that have the potential to stifle overall investment in Internet infrastructure and limit competition in telecommunications markets.

25 F.C.C.R. at 17970 ¶ 120 (emphasis added).

6

. In this case, Verizon has indicated it does wish to explore two-sided pricing (charging both edge providers and consumers).

7

. My research has not revealed any use of the phrase “terminating monopoly” outside of the context of these proceedings before the FCC. It does not appear to be an accepted economic term. A "gatekeeper,” on the other hand, is an intermediary between a consumer and an upstream seller. And a consumer’s willingness to switch to another available supplier depends on the prospective benefit measured against the transaction costs (how many blocks am I willing to walk, or how many phone calls am I willing to make?).

Recent literature suggests that gatekeepers may sometimes exercise market power against upstream suppliers even when the gatekeeper does not have enough market share to exercise downstream market power against consumers. See, e.g., Grimes, Warren S., Buyer Power and Retail Gatekeeper Power: Protecting Competition and the Atomistic Seller, 72 Antitrust L.J. 563, 580 (2005). One example would be if I purchase my groceries at a particular store, any food supplier who wishes to sell to me probably must do so through that particular store because I am unlikely to switch grocery stores over a single product. Regardless of any contemporary debates over the differences between buyer power and seller power, one thing is clear: The gatekeeper effect is a tool that facilitates the exercise of market power over sellers; it is not market power itself.

8

. The Commission treats each individual edge provider as analogous to an upstream seller in a retail context. But it seems more plausible that consumers consider "Internet access” to be the product that they are buying, and that large product creates greater incentives to switch to another provider. Although the Commission has argued that consumers will perceive a slow connection to a particular edge provider as indicative of a problem with that edge provider, rather than as a problem with the quality of Internet access provided by the broadband provider, 25 F.C.C.R. at 17921 ¶27, the Commission presents no evidence to support that conclusion. Indeed, edge providers have a strong incentive to inform consumers if their connections are being degraded. Moreover, the transparency rule, which we uphold, makes this outcome almost impossible.

9

. I do think that the transparency rules rest on firmer ground. The Commission is required to make triennial reports to Congress on "market entry barriers" in information services, 47 U.S.C. § 257, and requiring disclosure of network management practices appears to be reasonably ancillary to that duty. I also agree with the majority's conclusion that the disclosure rules are severable from the anti-discrimination and anti-blocking rules.

No. 81-1687.

SONY CORPORATION OF AMERICA et al. v. UNIVERSAL CITY STUDIOS, INC., et al.

Decided January 17, 1984

Argued January 18, 1983 — Reargued October 3, 1983

with whom Justice Marshall, Justice Powell, and Justice Rehnquist join,

Dean C. Dunlavey reargued the cause for petitioners. With him on the briefs were Donald E. Sloan and Marshall Rutter.

Stephen A. Kroft reargued the cause for respondents. With him on the brief was Sondra E. Berchin.*

*

Briefs of amici curiae urging reversal were filed for the Virginia Citi­zens’ Consumer Council, Inc., et al. by William A. Dobrovir; for the Amer­ican Library Association by Newton N. Minow; for the Consumer Elec­tronics Group by J. Edward Day; for the Educators Ad Hoc Committee on Copyright Law by Michael H. Cardozo, August W. Steinhilber, and Gwen­dolyn H. Gregory; for General Electric Co. et al. by Alfred B. Engelberg, Morton Amster, Jesse Rothstein, and Joel E. Lutzker; for Hitachi, Ltd., et al. by John W. Armagost and Craig B. Jorgensen; for McCann-­Erickson, Inc., et al. by John A. Donovan, A. Howard Matz, and David Fleischer; for Minnesota Mining and Manufacturing Co. et al. by Sidney A. Diamond and Grier Curran Raclin; for the National Retail Merchants As­sociation by Peter R. Stem, TheodoreS. Steingut, and Robert A. Weiner; for Sanyo Electric, Inc., by Anthony Liebig; for Sears, Roebuck and Co. by Max L. Gillam and Mary E. Woytek; for TDK Electronics Co., Ltd., by Ko-Yung Tung and Adam Yarmolinsky; for Toshiba Corp. et al. by Don­ald J. Zoeller and Herve Gouraige; for Pfizer Inc. by Steven C. Kany; and for Viare Publishing by Peter F. Marvin.

Briefs of amici curiae urging affirmance were filed for the Association of American Publishers, Inc., et al. by Charles H. Lieb and Jon A. Baum-­garten; for the Authors League of America, Inc., by Irwin Karp; for CBS Inc. by Lloyd N. Cutler, Louis R. Cohen, and George Vradenburg III; for Creators and Distributors of Programs by Stuart Robinowitz and Andrew J. Peck; for the International Alliance of Theatrical Stage Employees and Moving Picture Machine Operators of the United States and Canada, AFL-CIO, by Leo Geffner; for the Motion Picture Association of America, Inc., by Richard M. Cooper, Ellen S. Huvelle, and William Nix; for the National Music Publishers’ Association, Inc., by Jon A. Baumgarten; for the Recording Industry Association of America, Inc., by James F. Fitzpat­rick, Cary H. Sherman, and Ernest S. Meyers; for Volunteer Lawyers for the Arts, Inc., by I. Fred Koenigsberg; and for the Writers Guild of Amer­ica, West, Inc., et al. by Paul P. Selvin, Jerome B. Lurie, and Paul S. Berger.

Briefs of amici curiae were filed for the State of Missouri et al. by John Ashcroft, Attorney General of Missouri, and by the Attorneys General for their respective States as follows: Charles A. Graddick of Alabama, John Steven Clark of Arkansas, Michael J. Bowers of Georgia, Tany S. Hong of Hawaii, Tyrone C. Fahner of Illinois, Thomas J. Miller of Iowa, William J. Guste, Jr., oí Louisiana, William A. Attain of Mississippi, Michael T. Greely oí Montana, Rufus L. Edmisten of North Carolina, William J. Brown of Ohio, Jan Eric Cartwright of Oklahoma, Dennis J. Roberts II of Rhode Island, John J. Easton of Vermont, Gerald L. Battles of Virginia, and Bronson C. La Follette of Wisconsin; and for the Committee on Copy­right and Literary Property of the Association of the Bar of the City of New York by Michael S. Oberman and David H. Marks.

Justice Stevens

delivered the opinion of the Court.

Petitioners manufacture and sell home video tape record­ers. Respondents own the copyrights on some of the tele­vision programs that are broadcast on the public airwaves. Some members of the general public use video tape recorders sold by petitioners to record some of these broadcasts, as well as a large number of other broadcasts. The question presented is whether the sale of petitioners’ copying equip­ment to the general public violates any of the rights con­ferred upon respondents by the Copyright Act.

Respondents commenced this copyright infringement ac­tion against petitioners in the United States District Court for the Central District of California in 1976. Respondents alleged that some individuals had used Betamax video tape recorders (VTR’s) to record some of respondents’ copy­righted works which had been exhibited on commercially sponsored television and contended that these individuals had thereby infringed respondents’ copyrights. Respond­ents further maintained that petitioners were liable for the copyright infringement allegedly committed by Betamax con­sumers because of petitioners’ marketing of the Betamax VTR’s.1 Respondents sought no relief against any Beta-­max consumer. Instead, they sought money damages and an equitable accounting of profits from petitioners, as well as an injunction against the manufacture and marketing of Betamax VTR’s.

After a lengthy trial, the District Court denied respond­ents all the relief they sought and entered judgment for peti­tioners. 480 F. Supp. 429 (1979). The United States Court of Appeals for the Ninth Circuit reversed the District Court’s judgment on respondents’ copyright claim, holding petition­ers liable for contributory infringement and ordering the District Court to fashion appropriate relief. 659 F. 2d 963 (1981). We granted certiorari, 457 U. S. 1116 (1982); since we had not completed our study of the case last Term, we or­dered reargument, 463 U. S. 1226 (1983). We now reverse.

An explanation of our rejection of respondents’ unprece­dented attempt to impose copyright liability upon the distrib­utors of copying equipment requires a quite detailed recita­tion of the findings of the District Court. In summary, those findings reveal that the average member of the public uses a VTR principally to record a program he cannot view as it is being televised and then to watch it once at a later time. This practice, known as “time-shifting,” enlarges the tele­vision viewing audience. For that reason, a significant amount of television programming may be used in this man­ner without objection from the owners of the copyrights on the programs. For the same reason, even the two respond­ents in this case, who do assert objections to time-shifting in this litigation, were unable to prove that the practice has im­paired the commercial value of their copyrights or has cre­ated any likelihood of future harm. Given these findings, there is no basis in the Copyright Act upon which respond­ents can hold petitioners liable for distributing VTR’s to the general public. The Court of Appeals’ holding that respond­ents are entitled to enjoin the distribution of VTR’s, to collect royalties on the sale of such equipment, or to obtain other relief, if affirmed, would enlarge the scope of respondents’ statutory monopolies to encompass control over an article of commerce that is not the subject of copyright protection. Such an expansion of the copyright privilege is beyond the limits of the grants authorized by Congress.

The two respondents m this action, Universal City Studios, Inc., and Walt Disney Productions, produce and hold the copyrights on a substantial number of motion pictures and other audiovisual works. In the current marketplace, they can exploit their rights in these works in a number of ways: by authorizing theatrical exhibitions, by licensing limited showings on cable and network television, by selling syn­dication rights for repeated airings on local television sta­tions, and by marketing programs on prerecorded videotapes or videodiscs. Some works are suitable for exploitation through all of these avenues, while the market for other works is more limited.

Petitioner Sony manufactures millions of Betamax video tape recorders and markets these devices through numerous retail establishments, some of which are also petitioners in this action.2 Sony’s Betamax VTR is a mechanism consisting of three basic components: (1) a tuner, which receives elec­tromagnetic signals transmitted over the television band of the public airwaves and separates them into audio and visual signals; (2) a recorder, which records such signals on a mag­netic tape; and (3) an adapter, which converts the audio and visual signals on the tape into a composite signal that can be received by a television set.

Several capabilities of the machine are noteworthy. The separate tuner in the Betamax enables it to record a broad­cast off one station while the television set is tuned to another channel, permitting the viewer, for example, to watch two simultaneous news broadcasts by watching one “live” and re­cording the other for later viewing. Tapes may be reused, and programs that have been recorded may be erased either before or after viewing. A timer in the Betamax can be used to activate and deactivate the equipment at predetermined times, enabling an intended viewer to record programs that are transmitted when he or she is not at home. Thus a per­son may watch a program at home in the evening even though it was broadcast while the viewer was at work during the afternoon. The Betamax is also equipped with a pause button and a fast-forward control. The pause button, when depressed, deactivates the recorder until it is released, thus enabling a viewer to omit a commercial advertisement from the recording, provided, of course, that the viewer is present when the program is recorded. The fast-forward control enables the viewer of a previously recorded program to run the tape rapidly when a segment he or she does not desire to see is being played back on the television screen.

The respondents and Sony both conducted surveys of the way the Betamax machine was used by several hundred owners during a sample period in 1978. Although there were some differences in the surveys, they both showed that the primary use of the machine for most owners was “time-­shifting” — the practice of recording a program to view it once at a later time, and thereafter erasing it. Time-shifting en­ables viewers to see programs they otherwise would miss because they are not at home, are occupied with other tasks, or are viewing a program on another station at the time of a broadcast that they desire to watch. Both surveys also showed, however, that a substantial number of interviewees had accumulated libraries of tapes.3 Sony’s survey indicated that over 80% of the interviewees watched at least as much regular television as they had before owning a Betamax.4 Re­spondents offered no evidence of decreased television viewing by Betamax owners.5

Sony introduced considerable evidence describing televi­sion programs that could be copied without objection from any copyright holder, with special emphasis on sports, reli­gious, and educational programming,. For example, their survey indicated that 7.3% of all Betamax use is to record sports events, and representatives of professional baseball, football, basketball, and hockey testified that they had no ob­jection to the recording of their televised events for home use.6

Respondents offered opinion evidence concerning the fu­ture impact of the unrestricted sale of VTR’s on the commer­cial value of their copyrights. The District Court found, however, that they had failed to prove any likelihood of future harm from the use of VTR’s for time-shifting. 480 F. Supp., at 469.

The District Court’s Decision

The lengthy trial of the case in the District Court con­cerned the private, home use of VTR’s for recording pro­grams broadcast on the public airwaves without charge to the viewer.7 No issue concerning the transfer of tapes to other persons, the use of home-recorded tapes for public perform­ances, or the copying of programs transmitted on pay or cable television systems was raised. See id., at 432-433, 442.

The District Court concluded that noncommercial home use recording of material broadcast over the public airwaves was a fair use of copyrighted works and did not constitute copy­right infringement. It emphasized the fact that the material was broadcast free to the public at large, the noncommercial character of the use, and the private character of the activity conducted entirely within the home. Moreover, the court found that the purpose of this use served the public interest in increasing access to television programming, an interest that “is consistent with the First Amendment policy of pro­viding the fullest possible access to information through the public airwaves. Columbia Broadcasting System, Inc. v. Democratic National Committee, 412 U. S. 94, 102.” Id., at 454.8 Even when an entire copyrighted work was recorded, the District Court regarded the copying as fair use “because there is no accompanying reduction in the market for ‘plain­tiff’s original work.’” Ibid.

As an independent ground of decision, the District Court also concluded that Sony could not be held liable as a con­tributory infringer even if the home use of a VTR was consid­ered an infringing use. The District Court noted that Sony had no direct involvement with any Betamax purchasers who recorded copyrighted works off the air. Sony’s advertising was silent on the subject of possible copyright infringement, but its instruction booklet contained the following statement:

“Television programs, films, videotapes and other ma­terials may be copyrighted. Unauthorized recording of such material may be contrary to the provisions of the United States copyright laws.” Id., at 436.

The District Court assumed that Sony had constructive knowledge of the probability that the Betamax machine would be used to record copyrighted programs, but found that Sony merely sold a “product capable of a variety of uses, some of them allegedly infringing.” Id., at 461. It reasoned:

“Selling a staple article of commerce — e. g., a type­writer, a recorder, a camera, a photocopying machine— technically contributes to any infringing use subse­quently made thereof, but this kind of ‘contribution,’ if deemed sufficient as a basis for liability, would expand the theory beyond precedent and arguably beyond judi­cial management.

“.. . Commerce would indeed be hampered if manufac­turers of staple items were held liable as contributory in-­fringers whenever they ‘constructively’ knew that some purchasers on some occasions would use their product for a purpose which a court later deemed, as a matter of first impression, to be an infringement.” Ibid.

Finally, the District Court discussed the respondents’ prayer for injunctive relief, noting that they had asked for an injunction either preventing the future sale of Betamax ma­chines, or requiring that the machines be rendered incapable of recording copyrighted works off the air. The court stated that it had “found no case in which the manufacturers, dis­tributors, retailers and advertisers of the instrument en­abling the infringement were sued by the copyright holders,” and that the request for relief in this case “is unique.” Id., at 465.

It concluded that an injunction was wholly inappropriate because any possible harm to respondents was outweighed by the fact that “the Betamax could still legally be used to record noncopyrighted material or material whose owners consented to the copying. An injunction would deprive the public of the ability to use the Betamax for this noninfringing off-the-air recording.” Id., at 468.

The Court of Appeals’ Decision

The Court of Appeals reversed the District Court’s judg­ment on respondents’ copyright claim. It did not set aside any of the District Court’s findings of fact. Rather, it con­cluded as a matter of law that the home use of a VTR was not a fair use because it was not a “productive use.”9 It there­fore held that it was unnecessary for plaintiffs to prove any harm to the potential market for the copyrighted works, but then observed that it seemed clear that the cumulative effect of mass reproduction made possible by VTR’s would tend to diminish the potential market for respondents’ works. 659 F. 2d, at 974.

On the issue of contributory infringement, the Court of Ap­peals first rejected the analogy to staple articles of commerce such as tape recorders or photocopying machines. It noted that such machines “may have substantial benefit for some purposes” and do not “even remotely raise copyright prob­lems.” Id., at 975. VTR’s, however, are sold “for the pri­mary purpose of reproducing television programming” and “[virtually all” such programming is copyrighted material. Ibid. The Court of Appeals concluded, therefore, that VTR’s were not suitable for any substantial noninfringing use even if some copyright owners elect not to enforce their rights.

The Court of Appeals also rejected the District Court’s re­liance on Sony’s lack of knowledge that home use constituted infringement. Assuming that the statutory provisions defin­ing the remedies for infringement applied also to the non-­statutory tort of contributory infringement, the court stated that a defendant’s good faith would merely reduce his dam­ages liability but would not excuse the infringing conduct. It held that Sony was chargeable with knowledge of the homeowner’s infringing activity because the reproduction of copyrighted materials was either “the most conspicuous use” or “the major use” of the Betamax product. Ibid.

On the matter of relief, the Court of Appeals concluded that “statutory damages may be appropriate” and that the District Court should reconsider its determination that an in­junction would not be an appropriate remedy; and, referring to “the analogous photocopying area,” suggested that a con­tinuing royalty pursuant to a judicially created compulsory license may very well be an acceptable resolution of the relief issue. Id., at 976.

II

Article I, § 8, of the Constitution provides:

“The Congress shall have Power ... To Promote the Progress of Science and useful Arts, by securing for lim­ited Times to Authors and Inventors the exclusive Right to their respective Writings and Discoveries.”

The monopoly privileges that Congress may authorize are neither unlimited nor primarily designed to provide a special private benefit. Rather, the limited grant is a means by which an important public purpose may be achieved. It is intended to motivate the creative activity of authors and in­ventors by the provision of a special reward, and to allow the public access to the products of their genius after the limited period of exclusive control has expired.

“The copyright law, like the patent statutes, makes re­ward to the owner a secondary consideration. In Fox Film Corp. v. Doyal, 286 U. S. 123, 127, Chief Justice Hughes spoke as follows respecting the copyright mo­nopoly granted by Congress, ‘The sole interest of the United States and the primary object in conferring the monopoly lie in the general benefits derived by the pub­lic from the labors of authors.’ It is said that reward to the author or artist serves to induce release to the public of the products of his creative genius.” United States v. Paramount Pictures, Inc., 334 U. S. 131, 158 (1948).

As the text of the Constitution makes plain, it is Congress that has been assigned the task of defining the scope of the limited monopoly that should be granted to authors or to in­ventors in order to give the public appropriate access to their work product. Because this task involves a difficult balance between the interests of authors and inventors in the control and exploitation of their writings and discoveries on the one hand, and society’s competing interest in the free flow of ideas, information, and commerce on the other hand, our pat­ent and copyright statutes have been amended repeatedly.10

From its beginning, the law of copyright has developed in response to significant changes in technology.11 Indeed, it was the invention of a new form of copying equipment — the printing press — that gave rise to the original need for copy­right protection.12 Repeatedly, as new developments have occurred in this country, it has been the Congress that has fashioned the new rules that new technology made necessary. Thus, long before the enactment of the Copyright Act of 1909, 35 Stat. 1075, it was settled that the protection given to copyrights is wholly statutory. Wheaton v. Peters, 8 Pet. 591, 661-662 (1834). The remedies for infringement “are only those prescribed by Congress.” Thompson v. Hub­bard, 131 U. S. 123, 151 (1889).

The judiciary’s reluctance to expand the protections af­forded by the copyright without explicit legislative guidance is a recurring theme. See, e. g., Teleprompter Corp. v. Columbia Broadcasting System, Inc., 415 U. S. 394 (1974); Fortnightly Corp. v. United Artists Television, Inc., 392 U. S. 390 (1968); White-Smith Music Publishing Co. v. Apollo Co., 209 U. S. 1 (1908); Williams & Wilkins Co. v. United States, 203 Ct. Cl. 74, 487 F. 2d 1345 (1973), aff’d by an equally divided Court, 420 U. S. 376 (1975). Sound policy, as well as history, supports our consistent deference to Congress when major technological innovations alter the market for copyrighted materials. Congress has the con­stitutional authority and the institutional ability to accom­modate fully the varied permutations of competing interests that are inevitably implicated by such new technology.

In a case like this, in which Congress has not plainly marked our course, we must be circumspect in construing the scope of rights created by a legislative enactment which never contemplated such a calculus of interests. In doing so, we are guided by Justice Stewart’s exposition of the correct approach to ambiguities in the law of copyright:

“The limited scope of the copyright holder’s statutory monopoly, like the limited copyright duration required by the Constitution, reflects a balance of competing claims upon the public interest: Creative work is to be encouraged and rewarded, but private motivation must ultimately serve the cause of promoting broad public availability of literature, music, and the other arts. The immediate effect of our copyright law is to secure a fair return for an ‘author’s’ creative labor. But the ultimate aim is, by this incentive, to stimulate artistic creativity for the general public good. ‘The sole interest of the United States and the primary object in conferring the monopoly,’ this Court has said, ‘lie in the general bene­fits derived by the public from the labors of authors.’ Fox Film Corp. v. Doyal, 286 U. S. 123, 127. See Ken­dall v. Winsor, 21 How. 322, 327-328; Grant v. Ray­mond, 6 Pet. 218, 241-242. When technological change has rendered its literal terms ambiguous, the Copyright Act must be construed in light of this basic purpose.” Twentieth Century Music Corp. v. Aiken, 422 U. S. 151, 156 (1975) (footnotes omitted).

Copyright protection “subsists ... in original works of authorship fixed in any tangible medium of expression.” 17 U. S. C. § 102(a) (1982 ed.). This protection has never ac­corded the copyright owner complete control over all possible uses of his work.13 Rather, the Copyright Act grants the copyright holder “exclusive” rights to use and to authorize the use of his work in five qualified ways, including repro­duction of the copyrighted work in copies. § 106.14 All re­productions of the work, however, are not within the exclu­sive domain of the copyright owner; some are in the public domain. Any individual may reproduce a copyrighted work for a “fair use”; the copyright owner does not possess the exclusive right to such a use. Compare § 106 with § 107.

“Anyone who violates any of the exclusive rights of the copyright owner,” that is, anyone who trespasses into his exclusive domain by using or authorizing the use of the copy­righted work in one of the five ways set forth in the statute, “is an infringer of the copyright.” § 501(a). Conversely, anyone who is authorized by the copyright owner to use the copyrighted work in a way specified in the statute or who makes a fair use of the work is not an infringer of the copy­right with respect to such use.

The Copyright Act provides the owner of a copyright with a potent arsenal of remedies against an infringer of his work, including an injunction to restrain the infringer from violat­ing his rights, the impoundment and destruction of all re­productions of his work made in violation of his rights, a recovery of his actual damages and any additional profits re­alized by the infringer or a recovery of statutory damages, and attorney’s fees. §§502-505.15

The two respondents in this case do not seek relief against the Betamax users who have allegedly infringed their copy­rights. Moreover, this is not a class action on behalf of all copyright owners who license their works for television broadcast, and respondents have no right to invoke whatever rights other copyright holders may have to bring infringe­ment actions based on Betamax copying of their works.16 As was made clear by their own evidence, the copying of the re­spondents’ programs represents a small portion of the total use of VTR’s. It is, however, the taping of respondents’ own copyrighted programs that provides them with standing to charge Sony with contributory infringement. To prevail, they have the burden of proving that users of the Betamax have infringed their copyrights and that Sony should be held responsible for that infringement.

III

The Copyright Act does not expressly render anyone liable for infringement committed by another. In contrast, the Patent Act expressly brands anyone who “actively induces infringement of a patent” as an infringer, 35 U. S. C. § 271(b), and further imposes liability on certain individuals labeled “contributory” infringers, § 271(c). The absence of such ex­press language in the copyright statute does not preclude the imposition of liability for copyright infringements on certain parties who have not themselves engaged in the infringing activity.17 For vicarious liability is imposed in virtually all areas of the law, and the concept of contributory infringe­ment is merely a species of the broader problem of identify­ing the circumstances in which it is just to hold one individual accountable for the actions of another.

Such circumstances were plainly present in Kalem Co. v. Harper Brothers, 222 U. S. 55 (1911), the copyright decision of this Court on which respondents place their principal reli­ance. In Kalem, the Court held that the producer of an un­authorized film dramatization of the copyrighted book Ben Hur was liable for his sale of the motion picture to jobbers, who in turn arranged for the commercial exhibition of the film. Justice Holmes, writing for the Court, explained:

“The defendant not only expected but invoked by ad­vertisement the use of its films for dramatic reproduc­tion of the story. That was the most conspicuous pur­pose for which they could be used, and the one for which especially they were made. If the defendant did not contribute to the infringement it is impossible to do so except by taking part in the final act. It is liable on principles recognized in every part of the law.” Id., at 62-63.

The use for which the item sold in Kalem had been “espe­cially” made was, of course, to display the performance that had already been recorded upon it. The producer had per­sonally appropriated the copyright owner’s protected work and, as the owner of the tangible medium of expression upon which the protected work was recorded, authorized that use by his sale of the film to jobbers. But that use of the film was not his to authorize: the copyright owner possessed the exclusive right to authorize public performances of his work. Further, the producer personally advertised the unau­thorized public performances, dispelling any possible doubt as to the use of the film which he had authorized.

Respondents argue that Kalem stands for the proposition that supplying the “means” to accomplish an infringing activ­ity and encouraging that activity through advertisement are sufficient to establish liability for copyright infringement. This argument rests on a gross generalization that cannot withstand scrutiny. The producer in Kalem did not merely provide the “means” to accomplish an infringing activity; the producer supplied the work itself, albeit in a new medium of expression. Sony in the instant case does not supply Betamax consumers with respondents’ works; respondents do. Sony supplies a piece of equipment that is generally capable of copying the entire range of programs that may be televised: those that are uncopyrighted, those that are copyrighted but may be copied without objection from the copyright holder, and those that the copyright holder would prefer not to have copied. The Betamax can be used to make authorized or unauthorized uses of copyrighted works, but the range of its potential use is much broader than the particular infringing use of the film Ben Hur involved in Kalem. Kalem does not support respondents’ novel theory of liability.

Justice Holmes stated that the producer had “contributed” to the infringement of the copyright, and the label “contribu­tory infringement” has been applied in a number of lower court copyright cases involving an ongoing relationship be­tween the direct infringer and the contributory infringer at the time the infringing conduct occurred. In such cases, as in other situations in which the imposition of vicarious lia­bility is manifestly just, the “contributory” infringer was in a position to control the use of copyrighted works by others and had authorized the use without permission from the copyright owner.18 This case, however, plainly does not fall in that category. The only contact between Sony and the users of the Betamax that is disclosed by this record occurred at the moment of sale. The District Court expressly found that “no employee of Sony, Sonam or DDBI had either direct involvement with the allegedly infringing activity or direct contact with purchasers of Betamax who recorded copy­righted works off-the-air.” 480 F. Supp., at 460. And it further found that “there was no evidence that any of the copies made by Griffiths or the other individual witnesses in this suit were influenced or encouraged by [Sony’s] adver­tisements.” Ibid.

If vicarious liability is to be imposed on Sony in this case, it must rest on the fact that it has sold equipment with con­structive knowledge of the fact that its customers may use that equipment to make unauthorized copies of copyrighted material. There is no precedent in the law of copyright for the imposition of vicarious liability on such a theory. The closest analogy is provided by the patent law cases to which it is appropriate to refer because of the historic kinship between patent law and copyright law.19

In the Patent Act both the concept of infringement and the concept of contributory infringement are expressly de­fined by statute.20 The prohibition against contributory in­fringement is confined to the knowing sale of a component especially made for use in connection with a particular pat­ent. There is no suggestion in the statute that one patentee may object to the sale of a product that might be used in con­nection with other patents. Moreover, the Act expressly provides that the sale of a “staple article or commodity of commerce suitable for substantial noninfringing use” is not contributory infringement. 35 U. S. C. § 271(c).

When a charge of contributory infringement is predicated entirely on the sale of an article of commerce that is used by the purchaser to infringe a patent, the public interest in ac­cess to that article of commerce is necessarily implicated. A finding of contributory infringement does not, of course, re­move the article from the market altogether; it does, how­ever, give the patentee effective control over the sale of that item. Indeed, a finding of contributory infringement is nor­mally the functional equivalent of holding that the disputed article is within the monopoly granted to the patentee.21

For that reason, in contributory infringement cases arising under the patent laws the Court has always recognized the critical importance of not allowing the patentee to extend his monopoly beyond the limits of his specific grant. These cases deny the patentee any right to control the distribution of unpatented articles unless they are “unsuited for any com­mercial noninfringing use.” Dawson Chemical Co. v. Rohm & Hass Co., 448 U. S. 176, 198 (1980). Unless a commodity “has no use except through practice of the patented method,” id., at 199, the patentee has no right to claim that its distri­bution constitutes contributory infringement. “To form the basis for contributory infringement the item must almost be uniquely suited as a component of the patented invention.” P. Rosenberg, Patent Law Fundamentals § 17.02[2] (2d ed. 1982). “[A] sale of an article which though adapted to an in­fringing use is also adapted to other and lawful uses, is not enough to make the seller a contributory infringer. Such a rule would block the wheels of commerce.” Henry v. A. B. Dick Co., 224 U. S. 1, 48 (1912), overruled on other grounds, Motion Picture Patents Co. v. Universal Film Mfg. Co., 243 U. S. 502, 517 (1917).

We recognize there are substantial differences between the patent and copyright laws. But in both areas the contribu­tory infringement doctrine is grounded on the recognition that adequate protection of a monopoly may require the courts to look beyond actual duplication of a device or publi­cation to the products or activities that make such duplication possible. The staple article of commerce doctrine must strike a balance between a copyright holder’s legitimate de­mand for effective — not merely symbolic — protection of the statutory monopoly, and the rights of others freely to engage in substantially unrelated areas of commerce. Accordingly, the sale of copying equipment, like the sale of other articles of commerce, does not constitute contributory infringement if the product is widely used for legitimate, unobjectionable purposes. Indeed, it need merely be capable of substantial noninfringing uses.

IV

The question is thus whether the Betamax is capable of commercially significant noninfringing uses. In order to resolve that question, we need not explore all the different potential uses of the machine and determine whether or not they would constitute infringement. Rather, we need only consider whether on the basis of the facts as found by the District Court a significant number of them would be non-­infringing. Moreover, in order to resolve this case we need not give precise content to the question of how much use is commercially significant. For one potential use of the Beta-­max plainly satisfies this standard, however it is understood: private, noncommercial time-shifting in the home. It does so both (A) because respondents have no right to prevent other copyright holders from authorizing it for their pro­grams, and (B) because the District Court’s factual findings reveal that even the unauthorized home time-shifting of respondents’ programs is legitimate fair use.

A. Authorized, Time-Shifting

Each of the respondents owns a large inventory of valuable copyrights, but in the total spectrum of television program­ming their combined market share is small. The exact per­centage is not specified, but it is well below 10%.22 If they were to prevail, the outcome of this litigation would have a significant impact on both the producers and the viewers of the remaining 90% of the programming in the Nation. No doubt, many other producers share respondents’ concern about the possible consequences of unrestricted copying. Nevertheless the findings of the District Court make it clear that time-shifting may enlarge the total viewing audience and that many producers are willing to allow private time-shifting to continue, at least for an experimental time period.23

The District Court found:

“Even if it were deemed that home-use recording of copyrighted material constituted infringement, the Beta-­max could still legally be used to record noncopyrighted material or material whose owners consented to the copying. An injunction would deprive the public of the ability to use the Betamax for this noninfringing off-the-­air recording.

“Defendants introduced considerable testimony at trial about the potential for such copying of sports, re­ligious, educational and other programming. This in­cluded testimony from representatives of the Offices of the Commissioners of the National Football, Basketball, Baseball and Hockey Leagues and Associations, the Ex­ecutive Director of National Religious Broadcasters and various educational communications agencies. Plaintiffs attack the weight of the testimony offered and also con­tend that an injunction is warranted because infringing uses outweigh noninfringing uses.

“Whatever the future percentage of legal versus il­legal home-use recording might be, an injunction which seeks to deprive the public of the very tool or article of commerce capable of some noninfringing use would be an extremely harsh remedy, as well as one unprecedented in copyright law.” 480 F. Supp., at 468.

Although the District Court made these statements in the context of considering the propriety of injunctive relief, the statements constitute a finding that the evidence concerning “sports, religious, educational and other programming” was sufficient to establish a significant quantity of broadcasting whose copying is now authorized, and a significant potential for future authorized copying. That finding is amply sup­ported by the record. In addition to the religious and sports officials identified explicitly by the District Court,24 two items in the record deserve specific mention.

First is the testimony of John Kenaston, the station man­ager of Channel 58, an educational station in Los Angeles affiliated with the Public Broadcasting Service. He ex­plained and authenticated the station’s published guide to its programs.25 For each program, the guide tells whether un­limited home taping is authorized, home taping is authorized subject to certain restrictions (such as erasure within seven days), or home taping is not authorized at all. The Spring 1978 edition of the guide described 107 programs. Sixty-two of those programs or 58% authorize some home taping. Twenty-one of them or almost 20% authorize unrestricted home taping.26

Second is the testimony of Fred Rogers, president of the corporation that produces and owns the copyright on Mister Rogers’ Neighborhood. The program is carried by more public television stations than any other program. Its audi­ence numbers over 3,000,000 families a day. He testified that he had absolutely no objection to home taping for non­commercial use and expressed the opinion that it is a real service to families to be able to record children’s programs and to show them at appropriate times.27

If there are millions of owners of VTR’s who make copies of televised sports events, religious broadcasts, and educational programs such as Mister Rogers’ Neighborhood, and if the proprietors of those programs welcome the practice, the busi­ness of supplying the equipment that makes such copying fea­sible should not be stifled simply because the equipment is used by some individuals to make unauthorized reproductions of respondents’ works. The respondents do not represent a class composed of all copyright holders. Yet a finding of con­tributory infringement would inevitably frustrate the inter­ests of broadcasters in reaching the portion of their audience that is available only through time-shifting.

Of course, the fact that other copyright holders may wel­come the practice of time-shifting does not mean that re­spondents should be deemed to have granted a license to copy their programs. Third-party conduct would be wholly irrel­evant in an action for direct infringement of respondents’ copyrights. But in an action for contributory infringement against the seller of copying equipment, the copyright holder may not prevail unless the relief that he seeks affects only his programs, or unless he speaks for virtually all copyright hold­ers with an interest in the outcome. In this case, the record makes it perfectly clear that there are many important pro­ducers of national and local television programs who find nothing objectionable about the enlargement in the size of the television audience that results from the practice of time-­shifting for private home use.28 The seller of the equipment that expands those producers’ audiences cannot be a con­tributory infringer if, as is true in this case, it has had no direct involvement with any infringing activity.

B. Unauthorized Time-Shifting

Even unauthorized uses of a copyrighted work are not nec­essarily infringing. An unlicensed use of the copyright is not an infringement unless it conflicts with one of the specific ex­clusive rights conferred by the copyright statute. Twentieth Century Music Corp. v. Aiken, 422 U. S., at 154-155. More­over, the definition of exclusive rights in § 106 of the present Act is prefaced by the words “subject to sections 107 through 118.” Those sections describe a variety of uses of copy­righted material that “are not infringements of copyright” “notwithstanding the provisions of section 106.” The most pertinent in this case is § 107, the legislative endorsement of the doctrine of “fair use.”29

That section identifies various factors30 that enable a court to apply an “equitable rule of reason” analysis to particular claims of infringement.31 Although not conclusive, the first factor requires that “the commercial or nonprofit character of an activity” be weighed in any fair use decision.32 If the Betamax were used to make copies for a commercial or profit-­making purpose, such use would presumptively be unfair. The contrary presumption is appropriate here, however, be­cause the District Court’s findings plainly establish that time-­shifting for private home use must be characterized as a noncommercial, nonprofit activity. Moreover, when one con­siders the nature of a televised copyrighted audiovisual work, see 17 U. S. C. § 107(2) (1982 ed.), and that time-shifting merely enables a viewer to see such a work which he had been invited to witness in its entirety free of charge, the fact that the entire work is reproduced, see § 107(3), does not have its ordinary effect of militating against a finding of fair use.33

This is not, however, the end of the inquiry because Con­gress has also directed us to consider “the effect of the use upon the potential market for or value of the copyrighted work.” §107(4). The purpose of copyright is to create incentives for creative effort. Even copying for noncom­mercial purposes may impair the copyright holder’s ability to obtain the rewards that Congress intended him to have. But a use that has no demonstrable effect upon the potential market for, or the value of, the copyrighted work need not be prohibited in order to protect the author’s incentive to create. The prohibition of such noncommercial uses would merely inhibit access to ideas without any countervailing benefit.34

Thus, although every commercial use of copyrighted mate­rial is presumptively an unfair exploitation of the monopoly privilege that belongs to the owner of the copyright, noncom­mercial uses are a different matter. A challenge to a non­commercial use of a copyrighted work requires proof either that the particular use is harmful, or that if it should become widespread, it would adversely affect the potential market for the copyrighted work. Actual present harm need not be shown; such a requirement would leave the copyright holder with no defense against predictable damage. Nor is it nec­essary to show with certainty that future harm will result. What is necessary is a showing by a preponderance of the evi­dence that some meaningful likelihood of future harm exists. If the intended use is for commercial gain, that likelihood may be presumed. But if it is for a noncommercial purpose, the likelihood must be demonstrated.

In this case, respondents failed to carry their burden with regard to home time-shifting. The District Court described respondents’ evidence as follows:

“Plaintiffs’ experts admitted at several points in the trial that the time-shifting without librarying would re­sult in ‘not a great deal of harm.’ Plaintiffs’ greatest concern about time-shifting is with ‘a point of important philosophy that transcends even commercial judgment.’ They fear that with any Betamax usage, ‘invisible bound­aries’ are passed: ‘the copyright owner has lost control over his program.’” 480 F. Supp., at 467.

Later in its opinion, the District Court observed:

“Most of plaintiffs’ predictions of harm hinge on specu­lation about audience viewing patterns and ratings, a measurement system which Sidney Sheinberg, MCA’s president, calls a ‘black art’ because of the significant level of imprecision involved in the calculations.” Id., at 469.35

There was no need for the District Court to say much about past harm. “Plaintiffs have admitted that no actual harm to their copyrights has occurred to date.” Id., at 451.

On the question of potential future harm from time-shifting, the District Court offered a more detailed analysis of the evidence. It rejected respondents’ “fear that persons ‘watch­ing’ the original telecast of a program will not be meas­ured in the live audience and the ratings and revenues will decrease,” by observing that current measurement technol­ogy allows the Betamax audience to be reflected. Id., at 466.36 It rejected respondents’ prediction “that live televi­sion or movie audiences will decrease as more people watch Betamax tapes as an alternative,” with the observation that “[tjhere is no factual basis for [the underlying] assumption.” Ibid}37 It rejected respondents’ “fear that time-shifting will reduce audiences for telecast reruns,” and concluded instead that “given current market practices, this should aid plain­tiffs rather than harm them.” Ibid.38 And it declared that respondents’ suggestion that “theater or film rental exhi­bition of a program will suffer because of time-shift recording of that program” “lacks merit.” Id., at 467.39

After completing that review, the District Court restated its overall conclusion several times, in several different ways. “Harm from time-shifting is speculative and, at best, mini­mal.” Ibid. “The audience benefits from the time-shifting capability have already been discussed. It is not implausible that benefits could also accrue to plaintiffs, broadcasters, and advertisers, as the Betamax makes it possible for more per­sons to view their broadcasts.” Ibid. “No likelihood of harm was shown at trial, and plaintiffs admitted that there had been no actual harm to date.” Id., at 468-469. “Testi­mony at trial suggested that Betamax may require adjust­ments in marketing strategy, but it did not establish even a likelihood of harm.” Id., at 469. “Television production by plaintiffs today is more profitable than it has ever been, and, in five weeks of trial, there was no concrete evidence to suggest that the Betamax will change the studios’ financial picture.” Ibid.

The District Court’s conclusions are buttressed by the fact that to the extent time-shifting expands public access to freely broadcast television programs, it yields societal bene­fits. In Community Television of Southern California v. Gottfried, 459 U. S. 498, 508, n. 12 (1983), we acknowledged the public interest in making television broadcasting more available. Concededly, that interest is not unlimited. But it supports an interpretation of the concept of “fair use” that requires the copyright holder to demonstrate some likelihood of harm before he may condemn a private act of time-shifting as a violation of federal law.

When these factors are all weighed in the “equitable rule of reason” balance, we must conclude that this record amply supports the District Court’s conclusion that home time-­shifting is fair use. In light of the findings of the Dis­trict Court regarding the state of the empirical data, it is clear that the Court of Appeals erred in holding that the statute as presently written bars such conduct.40

In summary, the record and findings of the District Court lead us to two conclusions. First, Sony demonstrated a sig­nificant likelihood that substantial numbers of copyright hold­ers who license their works for broadcast on free television would not object to having their broadcasts time-shifted by private viewers. And second, respondents failed to demon­strate that time-shifting would cause any likelihood of non-­minimal harm to the potential market for, or the value of, their copyrighted works. The Betamax is, therefore, capa­ble of substantial noninfringing uses. Sony’s sale of such equipment to the general public does not constitute contribu­tory infringement of respondents’ copyrights.

V

“The direction of Art. I is that Congress shall have the power to promote the progress of science and the useful arts. When, as here, the Constitution is permissive, the sign of how far Congress has chosen to go can come only from Congress.” Deepsouth Packing Co. v. Laitram Corp., 406 U. S. 518, 530 (1972).

One may search the Copyright Act in vain for any sign that the elected representatives of the millions of people who watch television every day have made it unlawful to copy a program for later viewing at home, or have enacted a flat prohibition against the sale of machines that make such copy­ing possible.

It may well be that Congress will take a fresh look at this new technology, just as it so often has examined other inno­vations in the past. But it is not our job to apply laws that have not yet been written. Applying the copyright statute, as it now reads, to the facts as they have been developed in this case, the judgment of the Court of Appeals must be reversed.

It is so ordered.

1

The respondents also asserted causes of action under state law and § 43(a) of the Trademark Act of 1946, 60 Stat. 441, 15 U. S. C. § 1125(a). These claims are not before this Court.

2

The four retailers are Carter Hawley Hales Stores, Inc., Associated Dry Goods Corp., Federated Department Stores, Inc., and Henry’s Cam­era Corp. The principal defendants are Sony Corporation, the manufac­turer of the equipment, and its wholly owned subsidiary, Sony Corporation of America. The advertising agency of Doyle Dane Bernback, Inc., also involved in marketing the Betamax, is also a petitioner. An individual VTR user, William Griffiths, was named as a defendant in the District Court, but respondents sought no relief against him. Griffiths is not a petitioner. For convenience, we shall refer to petitioners collectively as Sony.

3

As evidence of how a VTR may be used, respondents offered the tes­timony of William Griffiths. Griffiths, although named as an individual defendant, was a client of plaintiffs’ law firm. The District Court sum­marized his testimony as follows:

“He owns approximately 100 tapes. When Griffiths bought his Betamax, he intended not only to time-shift (record, play-back and then erase) but also to build a library of cassettes. Maintaining a library, however, proved too expensive, and he is now erasing some earlier tapes and reusing them.

“Griffiths copied about 20 minutes of a Universal motion picture called ‘Never Give An Inch,’ and two episodes from Universal television series entitled ‘Baa Baa Black Sheep’ and ‘Holmes and Yo Yo.’ He would have erased each of these but for the request of plaintiffs’ counsel that it be kept. Griffiths also testified that he had copied but already erased Universal films called ‘Alpha Caper’ (erased before anyone saw it) and ‘Amelia Ear­hart.’ At the time of his deposition Griffiths did not intend to keep any Universal film in his library.

“Griffiths has also recorded documentaries, news broadcasts, sporting events and political programs such as a rerun of the Nixon/Kennedy debate.” 480 F. Supp. 429, 436-437 (1979).

Four other witnesses testified to having engaged in similar activity.

4

The District Court summarized some of the findings in these surveys as follows:

“According to plaintiffs’ survey, 75.4% of the VTR owners use their machines to record for time-shifting purposes half or most of the time. Defendants’ survey showed that 96% of the Betamax owners had used the machine to record programs they otherwise would have missed.

“When plaintiffs asked interviewees how many cassettes were in their li­brary, 55.8% said there were 10 or fewer. In defendants’ survey, of the total programs viewed by interviewees in the past month, 70.4% had been viewed only that one time and for 57.9%, there were no plans for further viewing.” Id., at 438.

5

“81.9% of the defendants’ interviewees watched the same amount or more of regular television as they did before owning a Betamax. 83.2% reported their frequency of movie going was unaffected by Betamax.” Id., at 439.

6

See Defendants’ Exh. OT, Table 20; Tr. 2447-2450, 2480, 2486-2487, 2515-2516, 2530-2534.

7

The trial also briefly touched upon demonstrations of the Betamax by the retailer petitioners which were alleged to be infringements by respond­ents. The District Court held against respondents on this claim, 480 F. Supp., at 456-457, the Court of Appeals affirmed this holding, 659 F. 2d 963, 976 (1981), and respondents did not cross-petition on this issue.

8

The court also found that this “access is not just a matter of conven­ience, as plaintiffs have suggested. Access has been limited not simply by inconvenience but by the basic need to work. Access to the better pro­gram has also been limited by the competitive practice of counterprogram-­ming.” 480 F. Supp., at 454.

9

“Without a ‘productive use,’ i. e. when copyrighted material is repro­duced for its intrinsic use, the mass copying of the sort involved in this case precludes an application of fair use.” 659 F. 2d, at 971-972.

10

In its Report accompanying the comprehensive revision of the Copy­right Act in 1909, the Judiciary Committee of the House of Represent­atives explained this balance:

“The enactment of copyright legislation by Congress under the terms of the Constitution is not based upon any natural right that the author has in his writings, . . . but upon the ground that the welfare of the public will be served and progress of science and useful arts will be promoted by securing to authors for limited periods the exclusive rights to their writings. . . .

“In enacting a copyright law Congress must consider . . . two questions: First, how much will the legislation stimulate the producer and so benefit the public; and, second, how much will the monopoly granted be detri­mental to the public? The granting of such exclusive rights, under the proper terms and conditions, confers a benefit upon the public that out­weighs the evils of the temporary monopoly.” H. R. Rep. No. 2222, 60th Cong., 2d Sess., 7 (1909).

11

Thus, for example, the development and marketing of player pianos and perforated rolls of music, see White-Smith Music Publishing Co. v. Apollo Co., 209 U. S. 1 (1908), preceded the enactment of the Copyright Act of 1909; innovations in copying techniques gave rise to the statutory exemption for library copying embodied in § 108 of the 1976 revision of the copyright law; the development of the technology that made it possible to retransmit television programs by cable or by microwave systems, see Fortnightly Corp. v. United Artists Television, Inc., 392 U. S. 390 (1968), and Teleprompter Corp. v. Columbia Broadcasting System, Inc., 415 U. S. 394 (1974), prompted the enactment of the complex provisions set forth in 17 U. S. C. § 111(d)(2)(B) and § 111(d)(5) (1982 ed.) after years of detailed congressional study, see Eastern Microwave, Inc. v. Doubleday Sports, Inc., 691 F. 2d 125, 129 (CA2 1982).

By enacting the Sound Recording Amendment of 1971, 85 Stat. 391, Congress also provided the solution to the “record piracy” problems that had been created by the development of the audio tape recorder. Sony argues that the legislative history of that Act, see especially H. R. Rep. No. 92-487, p. 7 (1971), indicates that Congress did not intend to prohibit the private home use of either audio or video tape recording equipment. In view of our disposition of the contributory infringement issue, we ex­press no opinion on that question.

12

“Copyright protection became necessary with the invention of the printing press and had its early beginnings in the British censorship laws. The fortunes of the law of copyright have always been closely connected with freedom of expression, on the one hand, and with technological im­provements in means of dissemination, on the other. Successive ages have drawn different balances among the interest of the writer in the con­trol and exploitation of his intellectual property, the related interest of the publisher, and the competing interest of society in the untrammeled dis­semination of ideas.” Foreword to B. Kaplan, An Unhurried View of Copyright vii-viii (1967).

13

See, e. g., White-Smith Music Publishing Co. v. Apollo Co., 209 U. S., at 19; cf. Deep South Packing Co. v. Laitram Corp., 406 U. S. 518, 530-531 (1972). While the law has never recognized an author's right to absolute control of his work, the natural tendency of legal rights to express themselves in absolute terms to the exclusion of all else is particularly pro­nounced in the history of the constitutionally sanctioned monopolies of the copyright and the patent. See, e. g., United States v. Paramount Pic­tures, Inc., 334 U. S. 131, 156-158 (1948) (copyright owners claiming right to tie license of one film to license of another under copyright law); Fox Film Corp. v. Doyal, 286 U. S. 123 (1932) (copyright owner claiming copy­right renders it immune from state taxation of copyright royalties); Bobbs-­Merrill Co. v. Straus, 210 U. S. 339, 349-351 (1908) (copyright owner claiming that a right to fix resale price of his works within the scope of his copyright); International Business Machines Corp. v. United States, 298 U. S. 131 (1936) (patentees claiming right to tie sale of unpatented article to lease of patented device).

14

Section 106 of the Act provides:

“Subject to sections 107 through 118, the owner of copyright under this title has the exclusive rights to do and to authorize any of the following:

“(1) to reproduce the copyrighted work in copies or phonorecords;

“(2) to prepare derivative works based upon the copyrighted work;

“(3) to distribute copies or phonorecords of the copyrighted work to the public by sale or other transfer of ownership, or by rental, lease, or lending;

“(4) in the ease of literary, musical, dramatic, and choreographic works, pantomimes, and motion pictures and other audiovisual works, to perform the copyrighted work publicly; and

“(5) in the case of literary, musical, dramatic, and choreographic works, pantomimes, and pictorial, graphic, or sculptural works, including the indi­vidual images of a motion picture or other audiovisual work, to display the copyrighted work publicly.”

15

Moreover, anyone who willfully infringes the copyright to reproduce a motion picture for purposes of commercial advantage or private financial gain is subject to substantial criminal penalties, 17 U. S. C. § 506(a) (1982 ed.), and the fruits and instrumentalities of the crime are forfeited upon conviction, § 506(b).

16

In this regard, we reject respondents’ attempt to cast this action as comparable to a class action because of the positions taken by amici with copyright interests and their attempt to treat the statements made by amici as evidence in this case. See Brief for Respondents 1, and n. 1, 6, 52, 53, and n. 116. The stated desires of amici concerning the outcome of this or any litigation are no substitute for a class action, are not evidence in the case, and do not influence our decision; we examine an amicus curiae brief solely for whatever aid it provides in analyzing the legal questions before us.

17

As the District Court correctly observed, however, “the lines between direct infringement, contributory infringement and vicarious liability are not clearly drawn ... .” 480 F. Supp., at 457-458. The lack of clarity in this area may, in part, be attributable to the fact that an infringer is not merely one who uses a work without authorization by the copyright owner, but also one who authorizes the use of a copyrighted work without actual authority from the copyright owner.

We note the parties’ statements that the questions of Sony’s liability under the “doctrines” of “direct infringement” and “vicarious liability” are not nominally before this Court. Compare Brief for Respondents 9, n. 22, 41, n. 90, with Reply Brief for Petitioners 1, n. 2. We also observe, how­ever, that reasoned analysis of respondents’ unprecedented contributory infringement claim necessarily entails consideration of arguments and case law which may also be forwarded under the other labels, and indeed the parties to a large extent rely upon such arguments and authority in support of their respective positions on the issue of contributory infringement.

18

The so-called “dance hall cases,” Famous Music Corp. v. Bay State Harness Horse Racing & Breeding Assn., Inc., 554 F. 2d 1213 (CA1 1977) (racetrack retained infringer to supply music to paying customers); KECA Music, Inc. v. Dingus McGee’s Co., 432 F. Supp. 72 (WD Mo. 1977) (cocktail lounge hired musicians to supply music to paying customers); Dreamland Ball Room, Inc. v. Shapiro, Bernstein & Co., 36 F. 2d 354 (CA7 1929) (dance hall hired orchestra to supply music to paying custom­ers), are often contrasted with the so-called landlord-tenant cases, in which landlords who leased premises to a direct infringer for a fixed rental and did not participate directly in any infringing activity were found not to be liable for contributory infringement. E. g., Deutsch v. Arnold, 98 F. 2d 686 (CA2 1938).

In Shapiro, Bernstein & Co. v. H. L. Green Co., 316 F. 2d 304 (CA2 1963), the owner of 23 chainstores retained the direct infringer to run its record departments. The relationship was structured as a licensing ar­rangement, so that the defendant bore none of the business risk of running the department. Instead, it received 10% or 12% of the direct infringer’s gross receipts. The Court of Appeals concluded:

“[The dance-hall eases] and this one lie closer on the spectrum to the employer-employee model, than to the landlord-tenant model. . . . [0]n the particular facts before us, . . . Green’s relationship to its infringing licensee, as well as its strong concern for the financial success of the phono­graph record concession, renders it liable for the unauthorized sales of the ‘bootleg’ records.

“... [T]he imposition of vicarious liability in the ease before us cannot be deemed unduly harsh or unfair. Green has the power to police carefully the conduct of its concessionaire . . .; our judgment will simply encourage it to do so, thus placing responsibility where it can and should be effec­tively exercised." Id., at 308 (emphasis in original).

In Gershwin Publishing Corp. v. Columbia Artists Management, Inc., 443 F. 2d 1159 (CA2 1971), the direct infringers retained the contribu­tory infringer to manage their performances. The contributory infringer would contact each direct infringer, obtain the titles of the musical compo­sitions to be performed, print the programs, and then sell the programs to its own local organizations for distribution at the time of the direct infringe­ment. Id., at 1161. The Court of Appeals emphasized that the contribu­tory infringer had actual knowledge that the artists it was managing were performing copyrighted works, was in a position to police the infringing conduct of the artists, and derived substantial benefit from the actions of the primary infringers. Id., at 1163.

In Screen Gems-Columbia Music, Inc. v. Mark-Fi Records, Inc., 256 F. Supp. 399 (SDNY 1966), the direct infringer manufactured and sold boot­leg records. In denying a motion for summary judgment, the District Court held that the infringer’s advertising agency, the radio stations that advertised the infringer’s works, and the service agency that boxed and mailed the infringing goods could all be held liable, if at trial it could be demonstrated that they knew or should have known that they were dealing in illegal goods.

19

A. g., United States v. Paramount Pictures, Inc., 334 U. S., at 158; Fox Film Corp. v. Doyal, 286 U. S., at 131; Wheaton v. Peters, 8 Pet. 591, 657-658 (1834). The two areas of the law, naturally, are not identical twins, and we exercise the caution which we have expressed in the past in applying doctrine formulated in one area to the other. See generally Mazer v. Stein, 347 U. S. 201, 217-218 (1954); Bobbs-Merrill Co. v. Straus, 210 U. S., at 345.

We have consistently rejected the proposition that a similar kinship ex­ists between copyright law and trademark law, and in the process of doing so have recognized the basic similarities between copyrights and patents. The Trade-Mark Cases, 100 U. S. 82, 91-92 (1879); see also United Drug Co. v. Theodore Rectanus Co., 248 U. S. 90, 97 (1918) (trademark right “has little or no analogy” to copyright or patent); McLean v. Fleming, 96 U. S. 245, 254 (1878); Canal Co. v. Clark, 13 Wall. 311, 322 (1872). Given the fundamental differences between copyright law and trademark law, in this copyright case we do not look to the standard for contributory infringe­ment set forth in Inwood Laboratories, Inc. v. Ives Laboratories, Inc., 456 U. S. 844, 854-855 (1982), which was crafted for application in trademark cases. There we observed that a manufacturer or distributor could be held liable to the owner of a trademark if it intentionally induced a mer­chant down the chain of distribution to pass off its product as that of the trademark owner’s or if it continued to supply a product which could readily be passed off to a particular merchant whom it knew was mislabel­ing the product with the trademark owner’s mark. If Inwood’s narrow standard for contributory trademark infringement governed here, re­spondents’ claim of contributory infringement would merit little discussion. Sony certainly does not “intentionally induc[e]” its customers to make infringing uses of respondents’ copyrights, nor does it supply its products to identified individuals known by it to be engaging in continuing infringe­ment of respondents’ copyrights, see id., at 855.

20

Title 36 U. S. C. §271 provides:

“(a) Except as otherwise provided in this title, whoever without author­ity makes, uses or sells any patented invention, within the United States during the term of the patent therefor, infringes the patent.

“(b) Whoever actively induces infringement of a patent shall be liable as an infringer.

“(c) Whoever sells a component of a patented machine, manufacture, combination or composition, or a material or apparatus for use in practicing a patented process, constituting a material part of the invention, knowing the same to be especially made or especially adapted for use in an infringe­ment of such patent, and not a staple article or commodity of commerce suitable for substantial noninfringing use, shall be liable as a contributory infringer.

“(d) No patent owner otherwise entitled to relief for infringement or contributory infringement of a patent shall be denied relief or deemed guilty of misuse or illegal extension of the patent right by reason of his hav­ing done one or more of the following: (1) derived revenue from acts which if performed by another without his consent would constitute contributory infringement of the patent; (2) licensed or authorized another to perform acts which if performed without his consent would constitute contributory infringement of the patent; (3) sought to enforce his patent rights against infringement or contributory infringement.”

21

It seems extraordinary to suggest that the Copyright Act confers upon all copyright owners collectively, much less the two respondents in this case, the exclusive right to distribute VTR’s simply because they may be used to infringe copyrights. That, however, is the logical implication of their claim. The request for an injunction below indicates that respond­ents seek, in effect, to declare VTR’s contraband. Their suggestion in this Court that a continuing royalty pursuant to a judicially created compulsory license would be an acceptable remedy merely indicates that respondents, for their part, would be willing to license their claimed monopoly interest in VTR’s to Sony in return for a royalty.

22

The record suggests that Disney’s programs at the time of trial con­sisted of approximately one hour a week of network television and one syndicated series. Universal’s percentage in the Los Angeles market on commercial television stations was under 5%. See Tr. 532-533, 549-550.

23

The District Court did not make any explicit findings with regard to how much broadcasting is wholly uncopyrighted. The record does include testimony that at least one movie — My Man Godfrey — falls within that cat­egory, id., at 2300-2301, and certain broadcasts produced by the Federal Government are also uncopyrighted. See 17 U. S. C. § 105 (1982 ed.). Cf. Schnapper v. Foley, 215 U. S. App. D. C. 59, 667 F. 2d 102 (1981) (explaining distinction between work produced by the Government and work commissioned by the Government). To the extent such broadcasting is now significant, it further bolsters our conclusion. Moreover, since copyright protection is not perpetual, the number of audiovisual works in the public domain necessarily increases each year.

24

See Tr. 2447-2450 (Alexander Hadden, Major League Baseball); id., at 2480, 2486-2487 (Jay Moyer, National Football League); id., at 2515-­2516 (David Stern, National Basketball Association); id., at 2530-2534 (Gilbert Stein, National Hockey League); id., at 2543-2552 (Thomas Han­sen, National Collegiate Athletic Association); id., at 2565-2572 (Benjamin Armstrong, National Religious Broadcasters). Those officials were au­thorized to be the official spokespersons for their respective institutions in this litigation. Id., at 2432, 2479, 2509-2510, 2530, 2538, 2563. See Fed. Rule Civ. Proc. 30(b)(6).

25

Tr. 2863-2902; Defendants’ Exh. PI.

26

See also Tr. 2833-2844 (similar testimony by executive director of New Jersey Public Broadcasting Authority). Cf. id., at 2592-2605 (testimony by chief of New York Education Department’s Bureau of Mass Communi­cations approving home taping for educational purposes).

27

“Some public stations, as well as commercial stations, program the ‘Neighborhood’ at hours when some children cannot use it. I think that it’s a real service to families to be able to record such programs and show them at appropriate times. I have always felt that with the advent of all of this new technology that allows people to tape the ‘Neighborhood’ off-­the-air, and I’m speaking for the ‘Neighborhood’ because that’s what I produce, that they then become much more active in the programming of their family’s television life. Very frankly, I am opposed to people being programmed by others. My whole approach in broadcasting has always been ‘You are an important person just the way you are. You can make healthy decisions.’ Maybe I’m going on too long, but I just feel that any­thing that allows a person to be more active in the control of his or her life, in a healthy way, is important.” Id., at 2920-2921. See also Defendants’ Exh. PI, p. 85.

28

It may be rare for large numbers of copyright owners to authorize duplication of their works without demanding a fee from the copier. In the context of public broadcasting, however, the user of the copyrighted work is not required to pay a fee for access to the underlying work. The traditional method by which copyright owners capitalize upon the tele­vision medium — commercially sponsored free public broadcast over the public airwaves — is predicated upon the assumption that compensation for the value of displaying the works will be received in the form of advertising revenues.

In the context of television programming, some producers evidently be­lieve that permitting home viewers to make copies of their works off the air actually enhances the value of their copyrights. Irrespective of their rea­sons for authorizing the practice, they do so, and in significant enough numbers to create a substantial market for a noninfringing use of the Sony VTR’s. No one could dispute the legitimacy of that market if the produc­ers had authorized home taping of their programs in exchange for a license fee paid directly by the home user. The legitimacy of that market is not compromised simply because these producers have authorized home taping of their programs without demanding a fee from the home user. The copy­right law does not require a copyright owner to charge a fee for the use of his works, and as this record clearly demonstrates, the owner of a copy­right may well have economic or noneconomic reasons for permitting cer­tain kinds of copying to occur without receiving direct compensation from the copier. It is not the role of the courts to tell copyright holders the best way for them to exploit their copyrights: even if respondents’ competitors were ill-advised in authorizing home videotaping, that would not change the fact that they have created a substantial market for a paradigmatic noninfringing use of Sony’s product.

29

The Copyright Act of 1909, 35 Stat. 1075, did not have a “fair use” pro­vision. Although that Act’s compendium of exclusive rights “to print, reprint, publish, copy, and vend the copyrighted work” was broad enough to encompass virtually all potential interactions with a copyrighted work, the statute was never so construed. The courts simply refused to read the statute literally in every situation. When Congress amended the statute in 1976, it indicated that it “intended to restate the present judicial doc­trine of fair use, not to change, narrow, or enlarge it in any way.” H. R. Rep. No. 94-1476, p. 66 (1976).

30

Section 107 provides:

“Notwithstanding the provisions of section 106, the fair use of a copy­righted work, including such use by reproduction in copies or phonorecords or by any other means specified by that section, for purposes such as criti­cism, comment, news reporting, teaching (including multiple copies for classroom use), scholarship, or research, is not an infringement of copy­right. In determining whether the use made of a work in any particular case is a fair use the factors to be considered shall include—

“(1) the purpose and character of the use, including whether such use is of a commercial nature or is for nonprofit educational purposes;

“(2) the nature of the copyrighted work;

“(3) the amount and substantiality of the portion used in relation to the copyrighted work as a whole; and

“(4) the effect of the use upon the potential market for or value of the copyrighted work.” 17 U. S. C. § 107 (1982 ed.).

31

The House Report expressly stated that the fair use doctrine is an “equitable rule of reason” in its explanation of the fair use section:

“Although the courts have considered and ruled upon the fair use doc­trine over and over again, no real definition of the concept has ever emerged. Indeed, since the doctrine is an equitable rule of reason, no gen­erally applicable definition is possible, and each case raising the question must be decided on its own facts. . . .

“General intention behind the provision

“The statement of the fair use doctrine in section 107 offers some guid­ance to users in determining when the principles of the doctrine apply. However, the endless variety of situations and combinations of circum­stances that can rise in particular cases precludes the formulation of exact rules in the statute. The bill endorses the purpose and general scope of the judicial doctrine of fair use, but there is no disposition to freeze the doctrine in the statute, especially during a period of rapid technological change. Beyond a very broad statutory explanation of what fair use is and some of the criteria applicable to it, the courts must be free to adapt the doctrine to particular situations on a case-by-case basis.” H. R. Rep. No. 94-1476, swpra, at 65-66.

The Senate Committee similarly eschewed a rigid, bright-line approach to fair use. The Senate Report endorsed the view “that off-the-air record­ing for convenience” could be considered “fair use” under some circum­stances, although it then made it clear that it did not intend to suggest that off-the-air recording for convenience should be deemed fair use under any circumstances imaginable. S. Rep. No. 94-473, pp. 65-66 (1975). The latter qualifying statement is quoted by the dissent, post, at 481, and if read in isolation, would indicate that the Committee intended to con­demn all off-the-air recording for convenience. Read in context, however, it is quite clear that that was the farthest thing from the Committee’s intention.

32

“The Committee has amended the first of the criteria to be consid­ered — ‘the purpose and character of the use’ — to state explicitly that this factor includes a consideration of ‘whether such use is of a commercial nature or is for non-profit educational purposes.’ This amendment is not intended to be interpreted as any sort of not-for-profit limitation on edu­cational uses of copyrighted works. It is an express recognition that, as under the present law, the commercial or non-profit character of an activity, while not conclusive with respect to fair use, can and should be weighed along with other factors in fair use decisions.” H. R. Rep. No. 94-1476, supra, at 66.

33

It has been suggested that “consumptive uses of copyrights by home VTR users are commercial even if the consumer does not sell the home­made tape because the consumer will not buy tapes separately sold by the copyrighth older.” Home Recording of Copyrighted Works: Hearing be­fore the Subcommittee on Courts, Civil Liberties and the Administration of Justice of the House Committee on the Judiciary, 97th Cong., 2d Sess., pt. 2, p. 1250 (1982) (memorandum of Prof. Laurence H. Tribe). Further­more, “[t]he error in excusing such theft as noncommercial,” we are told, “can be seen by simple analogy: jewel theft is not converted into a noncom­mercial veniality if stolen jewels are simply worn rather than sold.” Ibid. The premise and the analogy are indeed simple, but they add nothing to the argument. The use to which stolen j ewelry is put is quite irrelevant in determining whether depriving its true owner of his present possessory in­terest in it is venial; because of the nature of the item and the true owner’s interests in physical possession of it, the law finds the taking objectionable even if the thief does not use the item at all. Theft of a particular item of personal property of course may have commercial significance, for the thief deprives the owner of his right to sell that particular item to any in­dividual. Time-shifting does not even remotely entail comparable con­sequences to the copyright owner. Moreover, the time-shifter no more steals the program by watching it once than does the live viewer, and the live viewer is no more likely to buy prerecorded videotapes than is the time-shifter. Indeed, no live viewer would buy a prerecorded videotape if he did not have access to a VTR.

34

Cf. A. Latman, Fair Use of Copyrighted Works (1958), reprinted in Study No. 14 for the Senate Committee on the Judiciary, Copyright Law Revision, Studies Prepared for the Subcommittee on Patents, Trademarks, and Copyrights, 86th Cong., 2d Sess., 30 (1960):

“In certain situations, the copyright owner suffers no substantial harm from the use of his work. . . . Here again, is the partial marriage between the doctrine of fair use and the legal maxim de minimus non curat lex.”

35

See also 480 F. Supp., at 451:

“It should be noted, however, that plaintiffs’ argument is more complicated and speculative than was the plaintiff’s in Williams & Wilkins. . . . Here, plaintiffs ask the court to find harm based on many more assumptions. . . . As is discussed more fully in Part IV infra, some of these assumptions are based on neither fact nor experience, and plaintiffs admit that they are to some extent inconsistent and illogical.”

36

«There was testimony at trial, however, that Nielsen Ratings has already developed the ability to measure when a Betamax in a sample home is recording the program. Thus, the Betamax owner will be meas­ured as a part of the live audience. The later diary can augment that measurement with information about subsequent viewing.” Id., at 466.

In a separate section, the District Court rejected plaintiffs’ suggestion that the commercial attractiveness of television broadcasts would be dimin­ished because Betamax owners would use the pause button or fast-forward control to avoid viewing advertisements:

“It must be remembered, however, that to omit commercials, Betamax owners must view the program, including the commercials, while record­ing. To avoid commercials during playback, the viewer must fast-forward and, for the most part, guess as to when the commercial has passed. For most recordings, either practice may be too tedious. As defendants’ sur­vey showed, 92% of the programs were recorded with commercials and only 25% of the owners fast-forward through them. Advertisers will have to make the same kinds of judgments they do now about whether persons viewing televised programs actually watch the advertisements which inter­rupt them.” Id., at 468.

37

“Here plaintiffs assume that people will view copies when they would otherwise be watching television or going to the movie theater. There is no factual basis for this assumption. It seems equally likely that Betamax owners will play their tapes when there is nothing on television they wish to see and no movie they want to attend. Defendants’ survey does not show any negative effect of Betamax ownership on television viewing or theater attendance.” Id., at 466.

38

“The underlying assumptions here are particularly difficult to accept. Plaintiffs explain that the Betamax increases access to the original tele­vised material and that the more people there are in this original audience, the fewer people the rerun will attract. Yet current marketing practices, including the success of syndication, show just the opposite. Today, the larger the audience for the original telecast, the higher the price plaintiffs can demand from broadcasters from rerun rights. There is no survey within the knowledge of this court to show that the rerun audience is com­prised of persons who have not seen the program. In any event, if ratings can reflect Betamax recording, original audiences may increase and, given market practices, this should aid plaintiffs rather than harm them.” Ibid.

39

“This suggestion lacks merit. By definition, time-shift recording en­tails viewing and erasing, so the program will no longer be on tape when the later theater run begins. Of course, plaintiffs may fear that the Beta-­max owners will keep the tapes long enough to satisfy all their interest in the program and will, therefore, not patronize later theater exhibitions. To the extent that this practice involves librarying, it is addressed in section V. C., infra. It should also be noted that there is no evidence to suggest that the public interest in later theatrical exhibitions of motion pictures will be reduced any more by Betamax recording than it already is by the television broadcast of the film.” Id., at 467.

40

The Court of Appeals chose not to engage in any “equitable rule of rea­son” analysis in this case. Instead, it assumed that the category of “fair use” is rigidly circumscribed by a requirement that every such use must be “productive.” It therefore concluded that copying a television program merely to enable the viewer to receive information or entertainment that he would otherwise miss because of a personal scheduling conflict could never be fair use. That understanding of “fair use” was erroneous.

Congress has plainly instructed us that fair use analysis calls for a sensi­tive balancing of interests. The distinction between “productive” and “un­productive” uses may be helpful in calibrating the balance, but it cannot be wholly determinative. Although copying to promote a scholarly endeavor certainly has a stronger claim to fair use than copying to avoid interrupting a poker game, the question is not simply two-dimensional. For one thing, it is not true that all copyrights are fungible. Some copyrights govern ma­terial with broad potential secondary markets. Such material may well have a broader claim to protection because of the greater potential for com­mercial harm. Copying a news broadcast may have a stronger claim to fair use than copying a motion picture. And, of course, not all uses are fungible. Copying for commercial gain has a much weaker claim to fair use than copying for personal enrichment. But the notion of social “pro­ductivity” cannot be a complete answer to this analysis. A teacher who copies to prepare lecture notes is clearly productive. But so is a teacher who copies for the sake of broadening his personal understanding of his specialty. Or a legislator who copies for the sake of broadening her under­standing of what her constituents are watching; or a constituent who copies a news program to help make a decision on how to vote.

Making a copy of a copyrighted work for the convenience of a blind per­son is expressly identified by the House Committee Report as an example of fair use, with no suggestion that anything more than a purpose to enter­tain or to inform need motivate the copying. In a hospital setting, using a VTR to enable a patient to see programs he would otherwise miss has no productive purpose other than contributing to the psychological well-being of the patient. Virtually any time-shifting that increases viewer access to television programming may result in a comparable benefit. The statu­tory language does not identify any dichotomy between productive and nonproductive time-shifting, but does require consideration of the eco­nomic consequences of copying.

Justice Blackmun,

dissenting.

A restatement of the facts and judicial history of this case is necessary, in my view, for a proper focus upon the issues. Respondents’ position is hardly so “unprecedented,” ante, at 421, in the copyright law, nor does it really embody a “gross generalization,” ante, at 436, or a “novel theory of liability,” ante, at 437, and the like, as the Court, in belittling their claims, describes the efforts of respondents.

I

The introduction of the home videotape recorder (VTR) upon the market has enabled millions of Americans to make recordings of television programs in their homes, for future and repeated viewing at their own convenience. While this practice has proved highly popular with owners of television sets and VTR’s, it understandably has been a matter of con­cern for the holders of copyrights in the recorded programs. A result is the present litigation, raising the issues whether the home recording of a copyrighted television program is an infringement of the copyright, and, if so, whether the manu­facturers and distributors of VTR’s are liable as contribu­tory infringers. I would hope that these questions ultimately will be considered seriously and in depth by the Congress and be resolved there, despite the fact that the Court’s decision today provides little incentive for congressional action. Our task in the meantime, however, is to resolve these issues as best we can in the light of ill-fitting existing copyright law.

It is no answer, of course, to refer to and stress, as the Court does, this Court’s “consistent deference to Congress” when­ever “major technological innovations” appear. Ante, at 431. Perhaps a better and more accurate description is that the Court has tended to evade the hard issues when they arise in the area of copyright law. I see no reason for the Court to be particularly pleased with this tradition or to continue it. Indeed, it is fairly clear from the legislative history of the 1976 Act that Congress meant to change the old pattern and enact a statute that would cover new technologies, as well as old.

II

In 1976, respondents Universal City Studios, Inc., and Walt Disney Productions (Studios) brought this copyright infringement action in the United States District Court for the Central District of California against, among others, peti­tioners Sony Corporation, a Japanese corporation, and Sony Corporation of America, a New York corporation, the manu­facturer and distributor, respectively, of the Betamax VTR. The Studios sought damages, profits, and a wide-ranging injunction against further sales or use of the Betamax or Betamax tapes.

The Betamax, like other VTR’s, presently is capable of re­cording television broadcasts off the air on videotape cassettes, and playing them back at a later time.1 Two kinds of Beta-­max usage are at issue here.2 The first is “time-shifting,” whereby the user records a program in order to watch it at a later time, and then records over it, and thereby erases the program, after a single viewing. The second is “library-­building,” in which the user records a program in order to keep it for repeated viewing over a longer term. Sony’s ad­vertisements, at various times, have suggested that Betamax users “record favorite shows” or “build a library.” Sony’s Betamax advertising has never contained warnings about copyright infringement, although a warning does appear in the Betamax operating instructions.

The Studios produce copyrighted “movies” and other works that they release to theaters and license for television broad­cast. They also rent and sell their works on film and on prerecorded videotapes and videodiscs. License fees for television broadcasts are set according to audience ratings, compiled by rating services that do not measure any play­backs of videotapes. The Studios make the serious claim that VTR recording may result in a decrease in their revenue from licensing their works to television and from marketing them in other ways.

After a 5-week trial, the District Court, with a detailed opinion, ruled that home VTR recording did not infringe the Studios’ copyrights under either the Act of Mar. 4,1909 (1909 Act), 35 Stat. 1075, as amended (formerly codified as 17 U. S. C. § 1 et seq.), or the Copyright Revision Act of 1976 (1976 Act), 90 Stat. 2541, 17 U. S. C. § 101 et seq. (1982 ed.).3 The District Court also held that even if home VTR recording were an infringement, Sony could not be held liable under theories of direct infringement, contributory infringement, or vicarious liability. Finally, the court concluded that an in­junction against sales of the Betamax would be inappropriate even if Sony were liable under one or more of those theories. 480 F. Supp. 429 (1979).

The United States Court of Appeals for the Ninth Circuit reversed in virtually every respect. 659 F. 2d 963 (1981). It held that the 1909 Act and the 1976 Act contained no im­plied exemption for “home use” recording, that such record­ing was not “fair use,” and that the use of the Betamax to record the Studios’ copyrighted works infringed their copy­rights. The Court of Appeals also held Sony liable for con­tributory infringement, reasoning that Sony knew and antici­pated that the Betamax would be used to record copyrighted material off the air, and that Sony, indeed, had induced, caused, or materially contributed to the infringing conduct. The Court of Appeals remanded the case to the District Court for appropriate relief; it suggested that the District Court could consider the award of damages or a continuing royalty in lieu of an injunction. Id., at 976.

III

The Copyright Clause of the Constitution, Art. I, § 8, cl. 8, empowers Congress “To promote the Progress of Science and useful Arts, by securing for limited Times to Authors and In­ventors the exclusive Right to their respective Writings and Discoveries.” This Nation’s initial copyright statute was passed by the First Congress. Entitled “An Act for the en­couragement of learning,” it gave an author “the sole right and liberty of printing, reprinting, publishing and vending” his “map, chart, book or books” for a period of 14 years. Act of May 31, 1790, § 1, 1 Stat. 124. Since then, as the technol­ogy available to authors for creating and preserving their writings has changed, the governing statute has changed with it. By many amendments, and by complete revisions in 1831, 1870, 1909, and 1976,4 authors’ rights have been expanded to provide protection to any “original works of authorship fixed in any tangible medium of expression,” in­cluding “motion pictures and other audiovisual works.” 17 U. S. C. § 102(a) (1982 ed.).5

Section 106 of the 1976 Act grants the owner of a copyright a variety of exclusive rights in the copyrighted work,6 includ­ing the right “to reproduce the copyrighted work in copies or phonorecords.”7 This grant expressly is made subject to §§ 107-118, which create a number of exemptions and limita­tions on the copyright owner’s rights. The most important of these sections, for present purposes, is § 107; that section states that “the fair use of a copyrighted work ... is not an infringement of copyright.”8

The 1976 Act, like its predecessors,9 does not give the copyright owner full and complete control over all possible uses of his work. If the work is put to some use not enumer­ated in §106, the use is not an infringement. See Fort­nightly Corp. v. United Artists Television, Inc., 392 U. S. 390, 393-395 (1968). Thus, before considering whether home videotaping comes within the scope of the fair use ex­emption, one first must inquire whether the practice appears to violate the exclusive right, granted in the first instance by § 106(1), “to reproduce the copyrighted work in copies or phonorecords.”

A

Although the word “copies” is in the plural in §106(1), there can be no question that under the Act the making of even a single unauthorized copy is prohibited. The Senate and House Reports explain: “The references to ‘copies or phonorecords,’ although in the plural, are intended here and throughout the bill to include the singular (1 U. S. C. § l).”10 S. Rep. No. 94-473, p. 58 (1975) (1975 Senate Report); H. R. Rep. No. 94-1476, p. 61 (1976) (1976 House Report). The Reports then describe the reproduction right established by §106(1):

“[T]he right ‘to reproduce the copyrighted work in copies or phonorecords’ means the right to produce a material object in which the work is duplicated, transcribed, imi­tated, or simulated in a fixed form from which it can be ‘perceived, reproduced, or otherwise communicated, either directly or with the aid of a machine or device.’ As under the present law, a copyrighted work would be infringed by reproducing it in whole or in any sub­stantial part, and by duplicating it exactly or by imita­tion or simulation.” 1975 Senate Report 58; 1976 House Report 61.

The making of even a single videotape recording at home falls within this definition; the VTR user produces a material ob­ject from which the copyrighted work later can be perceived. Unless Congress intended a special exemption for the making of a single copy for personal use, I must conclude that VTR recording is contrary to the exclusive rights granted by § 106(1).

The 1976 Act and its accompanying Reports specify in some detail the situations in which a single copy of a copy­righted work may be made without infringement concerns. Section 108(a), for example, permits a library or archives “to reproduce no more than one copy or phonorecord of a work” for a patron, but only under very limited conditions; an entire work, moreover, can be copied only if it cannot be obtained elsewhere at a fair price.11 § 108(e); see also § 112(a) (broad­caster may “make no more than one copy or phonorecord of a particular transmission program,” and only under certain conditions). In other respects, the making of single copies is permissible only within the limited confines of the fair use doctrine. The Senate Report, in a section headed “Single and multiple copying,” notes that the fair use doctrine would permit a teacher to make a single copy of a work for use in the classroom, but only if the work was not a “sizable” one such as a novel or treatise. 1975 Senate Report 63-64; accord, 1976 House Report 68-69, 71. Other situations in which the making of a single copy would be fair use are described in the House and Senate Reports.12 But neither the statute nor its legislative history suggests any intent to create a general exemption for a single copy made for per­sonal or private use.

Indeed, it appears that Congress considered and rejected the very possibility of a special private use exemption. The issue was raised early in the revision process, in one of the studies prepared for Congress under the supervision of the Copyright Office. A. Latman, Pair Use of Copyrighted Works (1958), reprinted in Study No. 14 for the Senate Com­mittee on the Judiciary, Copyright Law Revision, Studies Prepared for the Subcommittee on Patents, Trademarks, and Copyrights, 86th Cong., 2d Sess., 1 (1960) (Latman Fair Use Study). This study found no reported case supporting the existence of an exemption for private use, although it noted that “the purpose and nature of a private use, and in some cases the small amount taken, might lead a court to apply the general principles of fair use in such a way as to deny liabil­ity.” Id., at 12. After reviewing a number of foreign copy­right laws that contained explicit statutory exemptions for private or personal use, id., at 25, Professor Latman outlined several approaches that a revision bill could take to the gen­eral issue of exemptions and fair use. One of these was the adoption of particularized rules to cover specific situations, including “the field of personal use.” Id., at 33.13

Rejecting the latter alternative, the Register of Copyrights recommended that the revised copyright statute simply men­tion the doctrine of fair use and indicate its general scope. The Register opposed the adoption of rules and exemptions to cover specific situations,14 preferring, instead, to rely on the judge-made fair use doctrine to resolve new problems as they arose. See Register’s 1961 Report 25; Register’s Supplementary Report 27-28.

The Register’s approach was reflected in the first copy­right revision bills, drafted by the Copyright Office in 1964. These bills, like the 1976 Act, granted the copyright owner the exclusive right to reproduce the copyrighted work, sub­ject only to the exceptions set out in later sections. H. R. 11947/S. 3008, 88th Cong., 2d Sess., §5(a) (1964). The pri­mary exception was fair use, § 6, containing language virtu­ally identical to § 107 of the 1976 Act. Although the copy­right revision bills underwent change in many respects from their first introduction in 1964 to their final passage in 1976, these portions of the bills did not change.15 I can con­clude only that Congress, like the Register, intended to rely on the fair use doctrine, and not on a per se exemption for private use, to separate permissible copying from the impermissible.16

When Congress intended special and protective treatment for private use, moreover, it said so explicitly. One such explicit statement appears in § 106 itself. The copyright owner’s exclusive right to perform a copyrighted work, in contrast to his right to reproduce the work in copies, is lim­ited. Section 106(4) grants a copyright owner the exclusive right to perform the work “publicly,” but does not afford the owner protection with respect to private performances by others. A motion picture is “performed” whenever its images are shown or its sounds are made audible. § 101. Like “sing­[ing] a copyrighted lyric in the shower,” Twentieth Century Music Corp. v. Aiken, 422 U. S. 151, 155 (1975), watching television at home with one’s family and friends is now con­sidered a performance. 1975 Senate Report 59-60; 1976 House Report 63.17 Home television viewing nevertheless does not infringe any copyright — but only because § 106(4) contains the word “publicly.”18 See generally 1975 Senate Report 60-61; 1976 House Report 63-64; Register’s 1961 Report 29-30. No such distinction between public and pri­vate uses appears in § 106(l)’s prohibition on the making of copies.19

Similarly, an explicit reference to private use appears in § 108. Under that section, a library can make a copy for a patron only for specific types of private use: “private study, scholarship, or research.”20 §§ 108(d)(1) and (e)(1); see 37 CFR §201.14(b) (1983). Limits also are imposed on the ex­tent of the copying and the type of institution that may make copies, and the exemption expressly is made inapplicable to motion pictures and certain other types of works. § 108(h). These limitations would be wholly superfluous if an entire copy of any work could be made by any person for private use.21

B

The District Court in this case nevertheless concluded that the 1976 Act contained an implied exemption for “home-use recording.” 480 F. Supp., at 444-446. The court relied pri­marily on the legislative history of a 1971 amendment to the 1909 Act, a reliance that this Court today does not duplicate. Ante, at 430, n. 11. That amendment, however, was ad­dressed to the specific problem of commercial piracy of sound recordings. Act of Oct. 15, 1971, 85 Stat. 391 (1971 Amend­ment). The House Report on the 1971 Amendment, in a section entitled “Home Recording,” contains the following statement:

“In approving the creation of a limited copyright in sound recordings it is the intention of the Committee that this limited copyright not grant any broader rights than are accorded to other copyright proprietors under the existing title 17. Specifically, it is not the intention of the Committee to restrain the home recording, from broadcasts or from tapes or records, of recorded per­formances, where the home recording is for private use and with no purpose of reproducing or otherwise capital­izing commercially on it. This practice is common and unrestrained today, and the record producers and per­formers would be in no different position from that of the owners of copyright in recorded musical composi­tions over the past 20 years.” H. R. Rep. No. 92-487, p. 7 (1971) (1971 House Report).

Similar statements were made during House hearings on the bill22 and on the House floor,23 although not in the Senate proceedings. In concluding that these statements created a general exemption for home recording, the District Court, in my view, paid too little heed to the context in which the statements were made, and failed to consider the limited pur­pose of the 1971 Amendment and the structure of the 1909 Act.

Unlike television broadcasts and other types of motion pictures, sound recordings were not protected by copyright prior to the passage of the 1971 Amendment. Although the underlying musical work could be copyrighted, the 1909 Act provided no protection for a particular performer’s rendition of the work. Moreover, copyrighted musical works that had been recorded for public distribution were subject to a “com­pulsory license”: any person was free to record such a work upon payment of a 2-cent royalty to the copyright owner. § 1(e), 35 Stat. 1075-1076. While reproduction without pay­ment of the royalty was an infringement under the 1909 Act, damages were limited to three times the amount of the un­paid royalty. § 25(e), 35 Stat. 1081-1082; Shapiro, Bernstein & Co. v. Goody, 248 F. 2d 260, 262-263, 265 (CA2 1957), cert. denied, 355 U. S. 952 (1958). It was observed that the prac­tical effect of these provisions was to legalize record piracy. See S. Rep. No. 92-72, p. 4 (1971); 1971 House Report 2.

In order to suppress this piracy, the 1971 Amendment ex­tended copyright protection beyond the underlying work and to the sound recordings themselves. Congress chose, how­ever, to provide only limited protection: owners of copyright in sound recordings were given the exclusive right “[t]o re­produce [their works] and distribute [them] to the public.” 1971 Amendment, § 1(a), 85 Stat. 391 (formerly codified as 17 U. S. C. § 1(f)).24 This right was merely the right of com­mercial distribution. See 117 Cong. Rec. 34748-34749 (1971) (colloquy of Reps. Kazen and Kastenmeier) (“the bill pro­tects copyrighted material that is duplicated for commercial purposes only”).

Against this background, the statements regarding home recording under the 1971 Amendment appear in a very dif­ferent light. If home recording was “common and unre­strained” under the 1909 Act, see 1971 House Report 7, it was because sound recordings had no copyright protection and the owner of a copyright in the underlying musical work could collect no more than a 2-cent royalty plus 6 cents in damages for each unauthorized use. With so little at stake, it is not at all surprising that the Assistant Register “d[id] not see anybody going into anyone’s home and preventing this sort of thing.” 1971 House Hearings 23.

But the references to home sound recording in the 1971 Amendment’s legislative history demonstrate no congres­sional intent to create a generalized home-use exemption from copyright protection. Congress, having recognized that the 1909 Act had been unsuccessful in controlling home sound recording, addressed only the specific problem of com­mercial record piracy. To quote Assistant Register Ringer again, home use was “not what this legislation [was] ad­dressed to.” Id., at 22.25

While the 1971 Amendment narrowed the sound record­ings loophole in then existing copyright law, motion pictures and other audiovisual works have been accorded full copy­right protection since at least 1912, see Act of Aug. 24, 1912, 37 Stat. 488, and perhaps before, see Edison v. Lubin, 122 F. 240 (CA3 1903), appeal dism’d, 195 U. S. 625 (1904). Con­gress continued this protection in the 1976 Act. Unlike the sound recording rights created by the 1971 Amendment, the reproduction rights associated with motion pictures under § 106(1) are not limited to reproduction for public distribu­tion; the copyright owner's right to reproduce the work ex­ists independently, and the “mere duplication of a copy may constitute an infringement even if it is never distributed.” Register's Supplementary Report 16; see 1975 Senate Report 57 and 1976 House Report 61. Moreover, the 1976 Act was intended as a comprehensive treatment of all aspects of copy­right law. The Reports accompanying the 1976 Act, unlike the 1971 House Report, contain no suggestion that home-use recording is somehow outside the scope of this all-inclusive statute. It was clearly the intent of Congress that no addi­tional exemptions were to be implied.26

I therefore find in the 1976 Act no implied exemption to cover the home taping of television programs, whether it be for a single copy, for private use, or for home use. Taping a copyrighted television program is infringement unless it is permitted by the fair use exemption contained in § 107 of the 1976 Act. I now turn to that issue.

IV

Fair Use

The doctrine of fair use has been called, with some justifi­cation, “the most troublesome in the whole law of copyright.” Dellar v. Samuel Goldwyn, Inc., 104 F. 2d 661, 662 (CA2 1939); see Triangle Publications, Inc. v. Knight-Ridder News­papers, Inc., 626 F. 2d 1171, 1174 (CA5 1980); Meeropol v. Nizer, 560 F. 2d 1061, 1068 (CA2 1977), cert. denied, 434 U. S. 1013 (1978). Although courts have constructed lists of factors to be considered in determining whether a particular use is fair,27 no fixed criteria have emerged by which that determination can be made. This Court thus far has pro­vided no guidance; although fair use issues have come here twice, on each occasion the Court was equally divided and no opinion was forthcoming. Williams & Wilkins Co. v. United States, 203 Ct. Cl. 74, 487 F. 2d 1345 (1973), aff’d, 420 U. S. 376 (1975); Benny v. Loew’s Inc., 239 F. 2d 532 (CA9 1956), aff’d sub nom. Columbia Broadcasting System, Inc. v. Loew’s Inc., 356 U. S. 43 (1958).

Nor did Congress provide definitive rules when it codified the fair use doctrine in the 1976 Act; it simply incorporated a list of factors “to be considered”: the “purpose and charac­ter of the use,” the “nature of the copyrighted work,” the “amount and substantiality of the portion used,” and, perhaps the most important, the “effect of the use upon the potential market for or value of the copyrighted work” (emphasis sup­plied). § 107. No particular weight, however, was assigned to any of these, and the list was not intended to be exclusive. The House and Senate Reports explain that §107 does no more than give “statutory recognition” to the fair use doc­trine; it was intended “to restate the present judicial doctrine of fair use, not to change, narrow, or enlarge it in any way.” 1976 House Report 66. See 1975 Senate Report 62; S. Rep. No. 93-983, p. 116 (1974); H. R. Rep. No. 83, 90th Cong., 1st Sess., 32 (1967); H. R. Rep. No. 2237, 89th Cong., 2d Sess., 61 (1966).

A

Despite this absence of clear standards, the fair use doc­trine plays a crucial role in the law of copyright. The pur­pose of copyright protection,, in the words of the Constitu­tion, is to “promote the Progress of Science and useful Arts.” Copyright is based on the belief that by granting authors the exclusive rights to reproduce their works, they are given an incentive to create, and that “encouragement of individual effort by personal gain is the best way to advance public welfare through the talents of authors and inventors in ‘Sci­ence and the useful Arts.’” Mazer v. Stein, 347 U. S. 201, 219 (1954). The monopoly created by copyright thus re­wards the individual author in order to benefit the public. Twentieth Century Music Corp. v. Aiken, 422 U. S., at 156; Fox Film Corp. v. Doyal, 286 U. S. 123, 127-128 (1932); see H. R. Rep. No. 2222, 60th Cong., 2d Sess., 7 (1909).

There are situations, nevertheless, in which strict enforce­ment of this monopoly would inhibit the very “Progress of Science and useful Arts” that copyright is intended to pro­mote. An obvious example is the researcher or scholar whose own work depends on the ability to refer to and to quote the work of prior scholars. Obviously, no author could create a new work if he were first required to repeat the re­search of every author who had gone before him.28 The scholar, like the ordinary user, of course could be left to bar­gain with each copyright owner for permission to quote from or refer to prior works. But there is a crucial difference be­tween the scholar and the ordinary user. When the ordinary user decides that the owner’s price is too high, and forgoes use of the work, only the individual is the loser. When the scholar forgoes the use of a prior work, not only does his own work suffer, but the public is deprived of his contribution to knowledge. The scholar’s work, in other words, produces external benefits from which everyone profits. In such a case, the fair use doctrine acts as a form of subsidy — albeit at the first author’s expense — to permit the second author to make limited use of the first author’s work for the public good. See Latman Fair Use Study 31; Gordon, Fair Use as Market Failure: A Structural Analysis of the Betamax Case and its Predecessors, 82 Colum. L. Rev. 1600, 1630 (1982).

A similar subsidy may be appropriate in a range of areas other than pure scholarship. The situations in which fair use is most commonly recognized are listed in § 107 itself; fair use may be found when a work is used “for purposes such as criti­cism, comment, news reporting, teaching, . . . scholarship, or research.” The House and Senate Reports expand on this list somewhat,29 and other examples may be found in the case law.30 Each of these uses, however, reflects a common theme: each is a productive use, resulting in some added ben­efit to the public beyond that produced by the first author’s work.31 The fair use doctrine, in other words, permits works to be used for “socially laudable purposes.” See Copyright Office, Briefing Papers on Current Issues, reprinted in 1975 House Hearings 2051, 2055. I am aware of no case in which the reproduction of a copyrighted work for the sole benefit of the user has been held to be fair use.32

I do not suggest, of course, that every productive use is a fair use. A finding of fair use still must depend on the facts of the individual case, and on whether, under the circum­stances, it is reasonable to expect the user to bargain with the copyright owner for use of the work. The fair use doc­trine must strike a balance between the dual risks created by the copyright system: on the one hand, that depriving au­thors of their monopoly will reduce their incentive to create, and, on the other, that granting authors a complete monopoly will reduce the creative ability of others.33 The inquiry is necessarily a flexible one, and the endless variety of situa­tions that may arise precludes the formulation of exact rules. But when a user reproduces an entire work and uses it for its original purpose, with no added benefit to the public, the doc­trine of fair use usually does not apply. There is then no need whatsoever to provide the ordinary user with a fair use subsidy at the author's expense.

The making of a videotape recording for home viewing is an ordinary rather than a productive use of the Studios’ copy­righted works. The District Court found that “Betamax owners use the copy for the same purpose as the original. They add nothing of their own.” 480 F. Supp., at 453. Al­though applying the fair use doctrine to home VTR record­ing, as Sony argues, may increase public access to material broadcast free over the public airwaves, I think Sony’s argu­ment misconceives the nature of copyright. Copyright gives the author a right to limit or even to cut off access to his work. Fox Film Corp. v. Doyal, 286 U. S., at 127. A VTR recording creates no public benefit sufficient to justify limit­ing this right. Nor is this right extinguished by the copy­right owner’s choice to make the work available over the airwaves. Section 106 of the 1976 Act grants the copy­right owner the exclusive right to control the performance and the reproduction of his work, and the fact that he has licensed a single television performance is really irrelevant to the existence of his right to control its reproduction. Al­though a television broadcast may be free to the viewer, this fact is equally irrelevant; a book borrowed from the public library may not be copied any more freely than a book that is purchased.

It may be tempting, as, in my view, the Court today is tempted, to stretch the doctrine of fair use so as to permit unfettered use of this new technology in order to increase ac­cess to television programming. But such an extension risks eroding the very basis of copyright law, by depriving authors of control over their works and consequently of their incen­tive to create.34 Even in the context of highly productive educational uses, Congress has avoided this temptation; in passing the 1976 Act, Congress made it clear that off-the-air videotaping was to be permitted only in very limited situa­tions. See 1976 House Report 71; 1975 Senate Report 64. And, the Senate Report adds, “[t]he committee does not in­tend to suggest. . . that off-the-air recording for convenience would under any circumstances, be considered ‘fair use.’” Id., at 66. I cannot disregard these admonitions.

B

I recognize, nevertheless, that there are situations where permitting even an unproductive use would have no effect on the author’s incentive to create, that is, where the use would not affect the value of, or the market for, the author’s work. Photocopying an old newspaper clipping to send to a friend may be an example; pinning a quotation on one’s bulletin board may be another. In each of these cases, the effect on the author is truly de minimis. Thus, even though these uses provide no benefit to the public at large, no purpose is served by preserving the author’s monopoly, and the use may be regarded as fair.

Courts should move with caution, however, in depriving authors of protection from unproductive “ordinary” uses. As has been noted above, even in the case of a productive use, § 107(4) requires consideration of “the effect of the use upon the potential market for or value of the copyrighted work” (emphasis added). “[A] particular use which may seem to have little or no economic impact on the author’s rights today can assume tremendous importance in times to come.” Register’s Supplementary Report 14. Although such a use may seem harmless when viewed in isolation, “[i]solated instances of minor infringements, when multiplied many times, become in the aggregate a major inroad on copy­right that must be prevented.” 1975 Senate Report 65.

I therefore conclude that, at least when the proposed use is an unproductive one, a copyright owner need prove only a potential for harm to the market for or the value of the copy­righted work. See 3 M. Nimmer, Copyright § 13.05[E][4][c], p. 13-84 (1983). Proof of actual harm, or even probable harm, may be impossible in an area where the effect of a new technology is speculative, and requiring such proof would present the “real danger ... of confining the scope of an au­thor’s rights on the basis of the present technology so that, as the years go by, his copyright loses much of its value because of unforeseen technical advances.” Register’s Supplemen­tary Report 14. Infringement thus would be found if the copyright owner demonstrates a reasonable possibility that harm will result from the proposed use. When the use is one that creates no benefit to the public at large, copyright pro­tection should not be denied on the basis that a new technol­ogy that may result in harm has not yet done so.

The Studios have identified a number of ways in which VTR recording could damage their copyrights. VTR re­cording could reduce their ability to market their works in movie theaters and through the rental or sale of prerecorded videotapes or videodiscs; it also could reduce their rerun au­dience, and consequently the license fees available to them for repeated showings. Moreover, advertisers may be will­ing to pay for only “live” viewing audiences, if they believe VTR viewers will delete commercials or if rating services are unable to measure VTR use; if this is the case, VTR record­ing could reduce the license fees the Studios are able to charge even for first-run showings. Library-building may raise the potential for each of the types of harm identified by the Studios, and time-shifting may raise the potential for substantial harm as well.35

Although the District Court found no likelihood of harm from VTR use, 480 F. Supp., at 468, I conclude that it ap­plied an incorrect substantive standard and misallocated the burden of proof. The District Court reasoned that the Stu­dios had failed to prove that library-building would occur “to any significant extent,” id., at 467; that the Studios’ prere­corded videodiscs could compete with VTR recordings and were “arguably . . . more desirable,” ibid.; that it was “not clear that movie audiences will decrease,” id., at 468; and that the practice of deleting commercials “may be too te­dious” for many viewers, ibid. To the extent any decrease in advertising revenues would occur, the court concluded that the Studios had “marketing alternatives at hand to recoup some of that predicted loss.” Id., at 452. Because the Stu­dios’ prediction of harm was “based on so many assumptions and on a system of marketing which is rapidly changing,” the court was “hesitant to identify ‘probable effects’ of home-use copying.” Ibid.

The District Court’s reluctance to engage in prediction in this area is understandable, but, in my view, the court was mistaken in concluding that the Studios should bear the risk created by this uncertainty. The Studios have demon­strated a potential for harm, which has not been, and could not be, refuted at this early stage of technological development.

The District Court's analysis of harm, moreover, failed to consider the effect of VTR recording on “the potential mar­ket for or the value of the copyrighted work,” as required by § 107(4).36 The requirement that a putatively infringing use of a copyrighted work, to be “fair,” must not impair a “poten­tial” market for the work has two implications. First, an infringer cannot prevail merely by demonstrating that the copyright holder suffered no net harm from the infringer’s ac­tion. Indeed, even a showing that the infringement has re­sulted in a net benefit to the copyright holder will not suffice. Rather, the infringer must demonstrate that he had not im­paired the copyright holder’s ability to demand compensation from (or to deny access to) any group who would otherwise be willing to pay to see or hear the copyrighted work. Second, the fact that a given market for a copyrighted work would not be available to the copyright holder were it not for the in-­fringer’s activities does not permit the infringer to exploit that market without compensating the copyright holder. See Iowa State University Research Foundation, Inc. v. American Broadcasting Cos., 621 F. 2d 57 (CA2 1980).

In this case, the Studios and their amici demonstrate that the advent of the VTR technology created a potential market for their copyrighted programs. That market consists of those persons who find it impossible or inconvenient to watch the programs at the time they are broadcast, and who wish to watch them at other times. These persons are willing to pay for the privilege of watching copyrighted work at their con­venience, as is evidenced by the fact that they are willing to pay for VTR’s and tapes; undoubtedly, most also would be willing to pay some kind of royalty to copyright holders. The Studios correctly argue that they have been deprived of the ability to exploit this sizable market.

It is thus apparent from the record and from the findings of the District Court that time-shifting does have a substantial adverse effect upon the “potential market for” the Studios’ copyrighted works. Accordingly, even under the formula­tion of the fair use doctrine advanced by Sony, time-shifting cannot be deemed a fair use.

V

Contributory Infringement

From the Studios’ perspective, the consequences of home VTR recording are the same as if a business had taped the Studios’ works off the air, duplicated the tapes, and sold or rented them to members of the public for home viewing. The distinction is that home VTR users do not record for commercial advantage; the commercial benefit accrues to the manufacturer and distributors of the Betamax. I thus must proceed to discuss whether the manufacturer and distribu­tors can be held contributorily liable if the product they sell is used to infringe.

It is well established that liability for copyright infringe­ment can be imposed on persons other than those who actu­ally carry out the infringing activity. Kalem Co. v. Harper Brothers, 222 U. S. 55, 62-63 (1911); 3 M. Nimmer, Copy­right § 12.04[A] (1983); see Twentieth Century Music Corp. v. Aiken, 422 U. S., at 160, n. 11; Buck v. Jewell-LaSalle Realty Co., 283 U. S. 191, 198 (1931). Although the liability provision of the 1976 Act provides simply that “[ajnyone who violates any of the exclusive rights of the copyright owner . . . is an infringer of the copyright,” 17 U. S. C. § 501(a) (1982 ed.), the House and Senate Reports demonstrate that Congress intended to retain judicial doctrines of contributory infringement. 1975 Senate Report 57; 1976 House Report 61.37

The doctrine of contributory copyright infringement, how­ever, is not well defined. One of the few attempts at defini­tion appears in Gershwin Publishing Corp. v. Columbia Art­ists Management, Inc., 443 F. 2d 1159 (CA2 1971). In that case the Second Circuit stated that “one who, with knowl­edge of the infringing activity, induces, causes or materially contributes to the infringing conduct of another, may be held liable as a ‘contributory’ infringer.” Id., at 1162 (footnote omitted). While I have no quarrel with this general state­ment, it does not easily resolve the present case; the District Court and the Court of Appeals, both purporting to apply it, reached diametrically opposite results.

A

In absolving Sony from liability, the District Court rea­soned that Sony had no direct involvement with individual Betamax users, did not participate in any off-the-air copying, and did not know that such copying was an infringement of the Studios’ copyright. 480 F. Supp., at 460. I agree with the Gershwin court that contributory liability may be im­posed even when the defendant has no formal control over the infringer. The defendant in Gershwin was a concert pro­moter operating through local concert associations that it sponsored; it had no formal control over the infringing per­formers themselves. 443 F. 2d, at 1162-1163. See also Twentieth Century Music Corp. v. Aiken, 422 U. S., at 160, n. 11. Moreover, a finding of contributory infringement has never depended on actual knowledge of particular instances of infringement; it is sufficient that the defendant have rea­son to know that infringement is taking place. 443 F. 2d, at 1162; see Screen Gems-Columbia Music, Inc. v. Mark-Fi Records, Inc., 256 F. Supp. 399 (SDNY 1966).38 In the so-­called “dance hall” cases, in which questions of contributory infringement arise with some frequency, proprietors of en­tertainment establishments routinely are held liable for un­authorized performances on their premises, even when they have no knowledge that copyrighted works are being per­formed. In effect, the proprietors in those cases are charged with constructive knowledge of the performances.39

Nor is it necessary that the defendant be aware that the infringing activity violates the copyright laws. Section 504(c)(2) of the 1976 Act provides for a reduction in statutory damages when an infringer proves he “was not aware and had no reason to believe that his or her acts constituted an infringement of copyright,” but the statute establishes no general exemption for those who believe their infringing activities are legal. Moreover, such an exemption would be meaningless in a case such as this, in which prospective relief is sought; once a court has established that the copying at issue is infringement, the defendants are necessarily aware of that fact for the future. It is undisputed in this case that Sony had reason to know the Betamax would be used by some owners to tape copyrighted works off the air. See 480 F. Supp., at 459-460.

The District Court also concluded that Sony had not caused, induced, or contributed materially to any infringing activities of Betamax owners. Id., at 460. In a case of this kind, however, causation can be shown indirectly; it does not depend on evidence that particular Betamax owners relied on particular advertisements. In an analogous case decided just two Terms ago, this Court approved a lower court’s con­clusion that liability for contributory trademark infringement could be imposed on a manufacturer who “suggested, even by implication” that a retailer use the manufacturer’s goods to infringe the trademark of another. Inwood Laboratories, Inc. v. Ives Laboratories, Inc., 456 U. S. 844, 851 (1982); see id., at 860 (opinion concurring in result). I think this stand­ard is equally appropriate in the copyright context.

The District Court found that Sony has advertised the Betamax as suitable for off-the-air recording of “favorite shows,” “novels for television,” and “classic movies,” 480 F. Supp., at 436, with no visible warning that such recording could constitute copyright infringement. It is only with the aid of the Betamax or some other VTR, that it is possible today for home television viewers to infringe copyright by recording off-the-air. Off-the-air recording is not only a foreseeable use for the Betamax, but indeed is its intended use. Under the circumstances, I agree with the Court of Appeals that if off-the-air recording is an infringement of copyright, Sony has induced and materially contributed to the infringing conduct of Betamax owners.40

B

Sony argues that the manufacturer or seller of a product used to infringe is absolved from liability whenever the prod­uct can be put to any substantial noninfringing use. Brief for Petitioners 41-42. The District Court so held, borrowing the “staple article of commerce” doctrine governing liability for contributory infringement of patents. See 85 U. S. C. §271.41 This Court today is much less positive. See ante, at 440-442. I do not agree that this technical judge-made doctrine of patent law, based in part on considerations irrele­vant to the field of copyright, see generally Dawson Chemi­cal Co. v. Rohm & Haas Co., 448 U. S. 176, 187-199 (1980), should be imported wholesale into copyright law. Despite their common constitutional source, see U. S. Const., Art. I, § 8, cl. 8, patent and copyright protections have not devel­oped in a parallel fashion, and this Court in copyright cases in the past has borrowed patent concepts only sparingly. See Bobbs-Merrill Co. v. Straus, 210 U. S. 339, 345-346 (1908).

I recognize, however, that many of the concerns underly­ing the “staple article of commerce” doctrine are present in copyright law as well. As the District Court noted, if liabil­ity for contributory infringement were imposed on the manu­facturer or seller of every product used to infringe — a type­writer, a camera, a photocopying machine — the “wheels of commerce” would be blocked. 480 F. Supp., at 461; see also Kalem Co. v. Harper Brothers, 222 U. S., at 62.

I therefore conclude that if a significant portion of the product’s use is noninfringing, the manufacturers and sellers cannot be held contributorily liable for the product’s infring­ing uses. See ante, at 440-441. If virtually all of the prod­uct’s use, however, is to infringe, contributory liability may be imposed; if no one would buy the product for noninfringing purposes alone, it is clear that the manufacturer is purposely profiting from the infringement, and that liability is appro­priately imposed. In such a case, the copyright owner’s monopoly would not be extended beyond its proper bounds; the manufacturer of such a product contributes to the infring­ing activities of others and profits directly thereby, while providing no benefit to the public sufficient to justify the infringement.

The Court of Appeals concluded that Sony should be held liable for contributory infringement, reasoning that “[video­tape recorders are manufactured, advertised, and sold for the primary purpose of reproducing television programming,” and “[virtually all television programming is copyrighted material.” 659 F. 2d, at 975. While I agree with the first of these propositions,42 the second, for me, is problematic. The key question is not the amount of television programming that is copyrighted, but rather the amount of VTR usage that is infringing.43 Moreover, the parties and their amici have argued vigorously about both the amount of television pro­gramming that is covered by copyright and the amount for which permission to copy has been given. The proportion of VTR recording that is infringing is ultimately a question of fact,44 and the District Court specifically declined to make findings on the “percentage of legal versus illegal home-use recording.” 480 F. Supp., at 468. In light of my view of the law, resolution of this factual question is essential. I there­fore would remand the case for further consideration of this by the District Court.

VI

The Court has adopted an approach very different from the one I have outlined. It is my view that the Court’s approach alters dramatically the doctrines of fair use and contributory infringement as they have been developed by Congress and the courts. Should Congress choose to respond to the Court’s decision, the old doctrines can be resurrected. As it stands, however, the decision today erodes much of the coherence that these doctrines have struggled to achieve.

The Court’s disposition of the case turns on its conclusion that time-shifting is a fair use. Because both parties agree that time-shifting is the primary use of VTR’s, that conclu­sion, if correct, would settle the issue of Sony’s liability under almost any definition of contributory infringement. The Court concludes that time-shifting is fair use for two reasons. Each is seriously flawed.

The Court’s first reason for concluding that time-shifting is fair use is its claim that many copyright holders have no ob­jection to time-shifting, and that “respondents have, no right to prevent other copyright holders from authorizing it for their programs.” Ante, at 442. The Court explains that a finding of contributory infringement would “inevitably frus­trate the interests of broadcasters in reaching the portion of their audience that is available only through time-shifting.” Ante, at 446. Such reasoning, however, simply confuses the question of liability with the difficulty of fashioning an appro­priate remedy. It may be that an injunction prohibiting the sale of VTR’s would harm the interests of copyright holders who have no objection to others making copies of their pro­grams. But such concerns should and would be taken into account in fashioning an appropriate remedy once liability has been found. Remedies may well be available that would not interfere with authorized time-shifting at all. The Court of Appeals mentioned the possibility of a royalty payment that would allow VTR sales and time-shifting to continue un­abated, and the parties may be able to devise other narrowly tailored remedies. Sony may be able, for example, to build a VTR that enables broadcasters to scramble the signal of individual programs and “jam” the unauthorized recording of them. Even were an appropriate remedy not available at this time, the Court should not misconstrue copyright holders’ rights in a manner that prevents enforcement of them when, through development of better techniques, an appropriate remedy becomes available.45

The Court’s second stated reason for finding that Sony is not liable for contributory infringement is its conclusion that even unauthorized time-shifting is fair use. Ante, at 447 et seq. This conclusion is even more troubling. The Court begins by suggesting that the fair use doctrine operates as a general “equitable rule of reason.” That interpretation mischaracterizes the doctrine, and simply ignores the lan­guage of the statute. Section 107 establishes the fair use doctrine “for purposes such as criticism, comment, news re­porting, teaching, . . . scholarship, or research.” These are all productive uses. It is true that the legislative history states repeatedly that the doctrine must be applied flexibly on a case-by-case basis, but those references were only in the context of productive uses. Such a limitation on fair use comports with its purpose, which is to facilitate the creation of new works. There is no indication that the fair use doc­trine has any application for purely personal consumption on the scale involved in this case,46 and the Court’s application of it here deprives fair use of the major cohesive force that has guided evolution of the doctrine in the past.

Having bypassed the initial hurdle for establishing that a use is fair, the Court then purports to apply to time-shifting the four factors explicitly stated in the statute. The first is “the purpose and character of the use, including whether such use is of a commercial nature or is for nonprofit educa­tional purposes.” §107(1). The Court confidently describes time-shifting as a noncommercial, nonprofit activity. It is clear, however, that personal use of programs that have been copied without permission is not what § 107(1) protects. The intent of the section is to encourage users to engage in activi­ties the primary benefit of which accrues to others. Time-­shifting involves no such humanitarian impulse. It is like­wise something of a mischaracterization of time-shifting to describe it as noncommercial in the sense that that term is used in the statute. As one commentator has observed, time-shifting is noncommercial in the same sense that steal­ing jewelry and wearing it — instead of reselling it — is non­commercial.47 Purely consumptive uses are certainly not what the fair use doctrine was designed to protect, and the awkwardness of applying the statutory language to time-­shifting only makes clearer that fair use was designed to pro­tect only uses that are productive.

The next two statutory factors are all but ignored by the Court — though certainly not because they have no applicabil­ity. The second factor — “the nature of the copyrighted work” — strongly supports the view that time-shifting is an infringing use. The rationale guiding application of this fac­tor is that certain types of works, typically those involving “more of diligence than of originality or inventiveness,” New York Times Co. v. Roxbury Data Interface, Inc., 434 P. Supp. 217, 221 (NJ 1977), require less copyright protection than other original works; Thus, for example, informational works, such as news reports, that readily lend themselves to productive use by others, are less protected than creative works of entertainment. Sony’s own surveys indicate that entertainment shows account for more than 80% of the pro­grams recorded by Betamax owners.48

The third statutory factor — “the amount and substanti-­ality of the portion used” — is even more devastating to the Court’s interpretation. It is undisputed that virtually all VTR owners record entire works, see 480 F. Supp., at 454, thereby creating an exact substitute for the copyrighted orig­inal. Fair use is intended to allow individuals engaged in productive uses to copy small portions of original works that will facilitate their own productive endeavors. Time-shifting bears no resemblance to such activity, and the complete du­plication that it involves might alone be sufficient to preclude a finding of fair use. It is little wonder that the Court has chosen to ignore this statutory factor.49

The fourth factor requires an evaluation of “the effect of the use upon the potential market for or value of the copy­righted work.” This is the factor upon which the Court focuses, but once again, the Court has misread the statute. As mentioned above, the statute requires a court to consider the effect of the use on the potential market for the copy­righted work. The Court has struggled mightily to show that VTR use has not reduced the value of the Studios’ copy­righted works in their present markets. Even if true, that showing only begins the proper inquiry. The development of the VTR has created a new market for the works produced by the Studios. That market consists of those persons who desire to view television programs at times other than when they are broadcast, and who therefore purchase VTR record­ers to enable them to time-shift.50 Because time-shifting of the Studios’ copyrighted works involves the copying of them, however, the Studios are entitled to share in the benefits of that new market. Those benefits currently go to Sony through Betamax sales. Respondents therefore can show harm from VTR use simply by showing that the value of their copyrights would increase if they were compensated for the copies that are used in the new market. The existence of this effect is self-evident.

Because of the Court’s conclusion concerning the legality of time-shifting, it never addresses the amount of noninfringing use that a manufacturer must show to absolve itself from liability as a contributory infringer. Thus, it is difficult to discuss how the Court’s test for contributory infringement would operate in practice under a proper analysis of time-­shifting. One aspect of the test as it is formulated by the Court, however, particularly deserves comment. The Court explains that a manufacturer of a product is not liable for con­tributory infringement as long as the product is “capable of substantial noninfringing uses.” Ante, at 442 (emphasis sup­plied). Such a definition essentially eviscerates the concept of contributory infringement. Only the most unimaginative manufacturer would be unable to demonstrate that a image-­duplicating product is “capable” of substantial noninfringing uses. Surely Congress desired to prevent the sale of prod­ucts that are used almost exclusively to infringe copyrights; the fact that noninfringing uses exist presumably would have little bearing on that desire.

More importantly, the rationale for the Court’s narrow standard of contributory infringement reveals that, once again, the Court has confused the issue of liability with that of remedy. The Court finds that a narrow definition of con­tributory infringement is necessary in order to protect “the rights of others freely to engage in substantially unrelated areas of commerce.” Ante, at 442. But application of the contributory infringement doctrine implicates such rights only if the remedy attendant upon a finding of liability were an injunction against the manufacture of the product in ques­tion. The issue of an appropriate remedy is not before the Court at this time, but it seems likely that a broad injunction is not the remedy that would be ordered. It is unfortunate that the Court has allowed its concern over a remedy to infect its analysis of liability.

VII

The Court of Appeals, having found Sony liable, remanded for the District Court to consider the propriety of injunctive or other relief. Because of my conclusion as to the issue of liability, I, too, would not decide here what remedy would be appropriate if liability were found. I concur, however, in the Court of Appeals’ suggestion that an award of damages, or continuing royalties, or even some form of limited injunc­tion, may well be an appropriate means of balancing the equi­ties in this case.51 Although I express no view on the merits of any particular proposal, I am certain that, if Sony were found liable in this case, the District Court would be able to fashion appropriate relief. The District Court might con­clude, of course, that a continuing royalty or other equitable relief is not feasible. The Studios then would be relegated to statutory damages for proven instances of infringement. But the difficulty of fashioning relief, and the possibility that complete relief may be unavailable, should not affect our interpretation of the statute.

Like so many other problems created by the interaction of copyright law with a new technology, “[t]here can be no really satisfactory solution to the problem presented here, until Congress acts.” Twentieth Century Music Corp. v. Aiken, 422 U. S., at 167 (dissenting opinion). But in the ab­sence of a congressional solution, courts cannot avoid difficult problems by refusing to apply the law. We must “take the Copyright Act... as we find it,” Fortnightly Corp. v. United Artists Television, Inc., 392 U. S., at 401-402, and “do as little damage as possible to traditional copyright principles . . . until the Congress legislates.” Id., at 404 (dissenting opinion).

1

The Betamax has three primary components: a tuner that receives tele­vision (“RF”) signals broadcast over the airwaves; an adapter that con­verts the RF signals into audio-video signals; and a recorder that places the audio-video signals on magnetic tape. Sony also manufactures VTR’s without built-in tuners; these are capable of playing back prerecorded tapes and recording home movies on videotape, but cannot record off the air. Since the Betamax has its own tuner, it can be used to record off one channel while another channel is being watched.

The Betamax is available with auxiliary features, including a timer, a pause control, and a fast-forward control; these allow Betamax owners to record programs without being present, to avoid (if they are present) recording commercial messages, and to skip over commercials while play­ing back the recording. Videotape is reusable; the user erases its record by recording over it.

2

This ease involves only the home recording for home use of television programs broadcast free over the airwaves. No issue is raised concerning cable or pay television, or the sharing or trading of tapes.

3

At the trial, the Studios proved 32 individual instances where their copyrighted works were recorded on Betamax VTR’s. Two of these in­stances occurred after January 1, 1978, the primary effective date of the 1976 Act; all the others occurred while the 1909 Act was still effective. My analysis focuses primarily on the 1976 Act, but the principles governing copyright protection for these works are the same under either Act.

4

Act of Feb. 3,1831, eh. 16, 4 Stat. 436; Act of July 8,1870, §§ 85-111,16 Stat. 212-217; Act of Mar. 4, 1909, 35 Stat. 1075 (formerly codified as 17 U. S. C. § 1 et seq.); Copyright Revision Act of 1976, 90 Stat. 2541 (codified as 17 U. S. C. § 101 et seq. (1982 ed.)).

5

Section 102(a) provides:

“Copyright protection subsists, in accordance with this title, in original works of authorship fixed in any tangible medium of expression, now known or later developed, from which they can be perceived, reproduced, or otherwise communicated, either directly or with the aid of a machine or device. Works of authorship include the following categories:

“(1) literary works;

“(2) musical works, including any accompanying words;

“(3) dramatic works, including any accompanying music;

“(4) pantomimes and choreographic works;

“(5) pictorial, graphic, and sculptural works;

“(6) motion pictures and other audiovisual works; and

“(7) sound recordings.”

Definitions of terms used in § 102(a)(6) are provided by § 101: “Audiovisual works” are “works that consist of a series of related images which are in­trinsically intended to be shown by the use of machines, or devices such as projectors, viewers, or electronic equipment, together -with accompanying sounds, if any, regardless of the nature of the material objects, such as films or tapes, in which the works are embodied.” And “motion pictures” are “audiovisual works consisting of a series of related images which, when shown in succession, impart an impression of motion, together with ac­companying sounds, if any.” Most commercial television programs, if fixed on film or tape at the time of broadcast or before, qualify as “audio­visual works.” Since the categories set forth in § 102(a) are not mutually exclusive, a particular television program may also qualify for protection as a dramatic, musical, or other type of work.

6

Section 106 provides:

“Subject to sections 107 through 118, the owner of copyright under this title has the exclusive rights to do and to authorize any of the following:

“(1) to reproduce the copyrighted work in copies or phonorecords;

“(2) to prepare derivative works based upon the copyrighted work;

“(3) to distribute copies or phonorecords of the copyrighted work to the public by sale or other transfer of ownership, or by rental, lease, or lending;

“(4) in the case of literary, musical, dramatic, and choreographic works, pantomimes, and motion pictures and other audiovisual works, to perform the copyrighted work publicly; and

“(5) in the case of literary, musical, dramatic, and choreographic works, pantomimes, and pictorial, graphic, or sculptural works, including the indi­vidual images of a motion picture or other audiovisual work, to display the copyrighted work publicly.”

7

A “phonorecord” is defined by § 101 as a reproduction of sounds other than sounds accompanying an audiovisual work, while a “copy” is a re­production of a work in any form other than a phonorecord.

8

Section 107 provides:

“Notwithstanding the provisions of section 106, the fair use of a copy­righted work, including such use by reproduction in copies or phonorecords or by any other means specified by that section, for purposes such as criti­cism, comment, news reporting, teaching (including multiple copies for classroom use), scholarship, or research, is not an infringement of copy­right. In determining whether the use made of a work in any particular case is a fair use the factors to be considered shall include—

“(1) the purpose and character of the use, including whether such use is of a commercial nature or is for nonprofit educational purposes;

“(2) the nature of the copyrighted work;

“(3) the amount and substantiality of the portion used in relation to the copyrighted work as a whole; and

“(4) the effect of the use upon the potential market for or value of the copyrighted work.”

Section 101 makes it clear that the four factors listed in this section are “illustrative and not limitative.”

9

The 1976 Act was the product of a revision effort lasting more than 20 years. Spurred by the recognition that “significant developments in technology and communications” had rendered the 1909 Act inadequate, S. Rep. No. 94-473, p. 47 (1975); see H. R. Rep. No. 94-1476, p. 47 (1976), Congress in 1955 authorized the Copyright Office to prepare a series of studies on all aspects of the existing copyright law. Thirty-four studies were prepared and presented to Congress. The Register of Copyrights drafted a comprehensive report with recommendations, House Committee on the Judiciary, Copyright Law Revision, Report of the Register of Copy­rights on the General Revision of the U. S. Copyright Law, 87th Cong., 1st Sess. (Comm. Print 1961) (Register’s 1961 Report), and general revision bills were introduced near the end of the 88th Congress in 1964. H. R. 11947/S. 3008, 88th Cong., 2d Sess. (1964). The Register issued a second report in 1965, with revised recommendations. House Committee on the Judiciary, Copyright Law Revision, pt. 6, Supplementary Report of the Register of Copyrights on the General Revision of the U. S. Copyright Law: 1965 Revision Bill, 89th Cong., 1st Sess. (Comm. Print 1965) (Regis­ter’s Supplementary Report). Action on copyright revision was delayed from 1967 to 1974 by a dispute on cable television, see generally Second Supplementary Report of the Register of Copyrights on the General Revi­sion of the U. S. Copyright Law: 1975 Revision Bill, ch. V, pp. 2-26 (Draft Oct.-Dec. 1975) (Register’s Second Supplementary Report), but a compro­mise led to passage of the present Act in 1976.

10

Title 1 U. S. C. § 1 provides in relevant part:

“In determining the meaning of any Act of Congress, unless the con­text indicates otherwise . . . words importing the plural include the singular . . . .”

11

The library photocopying provisions of § 108 do not excuse any person who requests “a copy” from a library if the requester’s use exceeds fair use. § 108(f)(2). Moreover, a library is absolved from liability for the un­supervised use of its copying equipment provided that the equipment bears a notice informing users that “the making of a copy” may violate the copy­right law. § 108(f)(1).

12

For example, “the making of a single copy or phonorecord by an indi­vidual as a free service for a blind person” would be a fair use, as would “a single copy reproduction of an excerpt from a copyrighted work by a callig­rapher for a single client” or “a single reproduction of excerpts from a copy­righted work by a student calligrapher or teacher in a learning situation.” 1975 Senate Report 66-67; see 1976 House Report 73-74. Application of the fair use doctrine in these situations, of eourse, would be unnecessary if the 1976 Act created a general exemption for the making of a single copy.

13

professor Latman made special mention of the “personal use” issue be­cause the area was one that “has become disturbed by recent developments .... Photoduplieation devices may make authors’ and publishers’ groups apprehensive. The Copyright Charter recently approved by [the Interna­tional Confederation of Societies of Authors and Composers] emphasizes the concern of authors over ‘private’ uses which, because of technological developments, are said to be competing seriously with the author’s eco­nomic interests.” Latman Fair Use Study 33-34.

14

The one exemption proposed by the Register, permitting a library to make a single photocopy of an out-of-print work and of excerpts that a requester certified were needed for research, met with opposition and was not included in the bills initially introduced in Congress. See Register’s 1961 Report 26; H. R. 11947/S. 3008, 88th Cong., 2d Sess. (1964); Regis­ter’s Supplementary Report 26. A library copying provision was restored to the bill in 1969, after pressure from library associations. Register’s Second Supplementary Report, ch. Ill, pp. 10-11; see S. 543, 91st Cong., 1st Sess., § 108 (Comm. Print, Dec. 10, 1969); 1975 Senate Report 48.

15

The 1964 bills provided that the fair use of copyrighted material for purposes “such as criticism, comment, news reporting, teaching, scholar­ship, or research” was not an infringement of copyright, and listed four “factors to be considered” in determining whether any other particular use was fair. H. R. 11947/S. 3008, 88th Cong., 2d Sess., § 6 (1964). Revised bills, drafted by the Copyright Office in 1965, contained a fair use provision merely mentioning the doctrine but not indicating its scope: “Notwith­standing the provisions of section 106, the fair use of a copyrighted work is not an infringement of copyright.” H. R. 4347/S. 1006, 89th Cong., 1st Sess., § 107 (1965). The House Judiciary Committee restored the provi­sion to its earlier wording, H. R. Rep. No. 2237, 89th Cong., 2d Sess., 5, 58 (1966), and the language adopted by the Committee remained in the bill in later Congresses. See H. R. 2512/S. 597, 90th Cong., 1st Sess., § 107 (1967); S. 543, 91st Cong., 1st Sess., §107 (1969); S. 644, 92d Cong., 1st Sess., §107 (1971); S. 1361, 93d Cong., 1st Sess., §107 (1973); H. R. 2223/S. 22, 94th Cong., 1st Sess., § 107 (1975). With a few additions by the House Judiciary Committee in 1976, see 1976 House Report 5; H. R. Conf. Rep. No. 94-1733, p. 70 (1976), the same language appears in § 107 of the 1976 Act.

16

In Williams & Wilkins Co. v. United States, 203 Ct. Cl. 74, 487 F. 2d 1345 (1973), aff’d by an equally divided Court, 420 U. S. 376 (1975), decided during the process of the revision of the copyright statutes, the Court of Claims suggested that copying for personal use might be outside the scope of copyright protection under the 1909 Act. The court reasoned that be­cause “hand copying” for personal use has always been regarded as permis­sible, and because the practice of making personal copies continued after typewriters and photostat machines were developed, the making of per­sonal copies by means other than hand copying should be permissible as well. 203 Ct. CL, at 84-88, 487 F. 2d, at 1350-1352.

There appear to me to be several flaws in this reasoning. First, it is by no means clear that the making of a “hand copy” of an entire work is permissible; the most that can be said is that there is no reported case on the subject, possibly because no copyright owner ever thought it worth­while to sue. See Latman Fair Use Study 11-12; 3 M. Nimmer, Copyright § 13.05[E][4][a] (1983). At least one early treatise asserted that infringe­ment would result “if an individual made copies for his personal use, even in his own handwriting, as there is no rule of law excepting manuscript copies from the law of infringement.” A. Weil, American Copyright Law § 1066 (1917). Second, hand copying or even copying by typewriter is self-­limiting. The drudgery involved in making hand copies ordinarily ensures that only necessary and fairly small portions of a work are taken; it is un­likely that any user would make a hand copy as a substitute for one that could be purchased. The harm to the copyright owner from hand copying thus is minimal. The recent advent of inexpensive and readily available copying machines, however, has changed the dimensions of the problem. See Register’s Second Supplementary Report, ch. Ill, p. 3; Hearings on H. R. 2223 before the Subcommittee on Courts, Civil Liberties, and the Administration of Justice of the House Committee on the Judiciary, 94th Cong., 1st Sess., 194 (1975) (1975 House Hearings) (remarks of Rep. Dan-­ielson); id., at 234 (statement of Robert W. Cairns); id., at 250 (remarks of Rep. Danielson); id., at 354 (testimony of Irwin Karp); id., at 467 (testi­mony of Rondo Cameron); id., at 1795 (testimony of Barbara Ringer, Reg­ister of Copyrights). Thus, “[t]he supposition that there is no tort in­volved in a scholar copying a copyrighted text by hand does not much advance the question of machine copying.” B. Kaplan, An Unhurried View of Copyright 101-102 (1967).

17

In a trio of eases, Fortnightly Corp. v. United Artists Television, Inc., 392 U. S. 390, 398 (1968); Teleprompter Corp. v. Columbia Broadcasting System, Inc., 415 U. S. 394, 403-405 (1974); and Twentieth Century Music Corp. v. Aiken, 422 U. S. 151 (1975), this Court had held that the reception of a radio or television broadcast was not a “performance” under the 1909 Act. The Court’s “narrow construction” of the word “perform” was “com­pletely overturned by the [1976 Act] and its broad definition of ‘perform’ in section 101.” 1976 House Report 87.

18

A work is performed “publicly” if it takes place “at a place open to the public or at any place where a substantial number of persons outside of a normal circle of a family and its social acquaintances is gathered.” § 101.

19

One purpose of the exemption for private performances was to permit the home viewing of lawfully made videotapes. The Register noted in 1961 that “[n]ew technical devices will probably make it practical in the future to reproduce televised motion pictures in the home. We do not be­lieve the private use of such a reproduction can or should be precluded by copyright.” Register’s 1961 Report 30 (emphasis added). The Register did not suggest that the private making of a reproduction of a televised motion picture would be permitted by the copyright law. The Register later reminded Congress that “[i]n general the concept of ‘performance’ must be distinguished sharply from the reproduction of copies.” Regis­ter’s Supplementary Report 22.

20

During hearings on this provision, Representative Danielson inquired whether it would apply to works of fiction such as “Gone With the Wind,” or whether it was limited to “strictly technical types of information.” The uncontradicted response was that it would apply only in “general terms of science . .. [and] the useful arts.” 1975 House Hearings 251 (testimony of Robert W. Cairns); cf. id., at 300 (statement of Harry Rosenfield) (“We are not asking ... for the right to copy ‘Gone With the Wind’ ”).

21

The mention in the Senate and House Reports of situations in which copies for private use would be permissible under the fair use doctrine — for example, the making of a free copy for a blind person, 1975 Senate Report 66; 1976 House Report 73, or the “recordings of performances by music stu­dents for purposes of analysis and criticism,” 1975 Senate Report 63— would be superfluous as well. See n. 12, supra.

22

The following exchange took place during the testimony of Barbara Ringer, then Assistant Register of Copyrights:

“[Rep.] Biester. ... I can tell you I must have a small pirate in my own home. My son has a cassette tape recorder, and as a particular record be­comes a hit, he will retrieve it onto his little set. . . . [T]his legislation, of course, would not point to his activities, would it?

“Miss Ringer. I think the answer is clearly,‘No, it would not.’ I have spoken at a couple of seminars on video cassettes lately, and this question is usually asked: ‘What about the home recorders?’ The answer I have given and will give again is that this is something you cannot control. You simply cannot control it. My own opinion, whether this is philosophical dogma or not, is that sooner or later there is going to be a crunch here. But that is not what this legislation is addressed to, and I do not see the crunch coming in the immediate future. ... I do not see anybody going into anyone’s home and preventing this sort of thing, or forcing legisla­tion that would engineer a piece of equipment not to allow home taping.” Hearings on S. 646 and H. R. 6927 before Subcommittee No. 3 of the House Committee on the Judiciary, 92d Cong., 1st Sess., 22-23 (1971) (1971 House Hearings).

23

Shortly before passage of the bill, a colloquy took place between Representative Kastenmeier, Chairman of the House Subcommittee that produced the bill, and Representative Kazen, who was not on the Subcommittee:

“Mr. KAZEN. Am I correct in assuming that the bill protects copy­righted material that is duplicated for commercial purposes only?

“Mr. KASTENMEIER. Yes.

“Mr. KAZEN. In other words, if your child were to record off of a pro­gram which comes through the air on the radio or television, and then used it for her own personal pleasure, for listening pleasure, this use would not be included under the penalties of this bill?

“Mr. KASTENMEIER. This is not included in the bill. I am glad the gentleman raises the point.

“On page 7 of the report, under ‘Home Recordings,’ Members will note that under the bill the same practice which prevails today is called for; namely, this is considered both presently and under the proposed law to be fair use. The child does not do this for commercial purposes. This is made clear in the report.” 117 Cong. Rec. 34748-34749 (1971).

24

The 1909 Act’s grant of an exclusive right to “copy,” § 1(a), was of no assistance to the owner of a copyright in a sound recording, because a reproduction of a sound recording was technically considered not to be a “copy.” See 1971 House Hearings 18 (testimony of Barbara Ringer, Assistant Register of Copyrights); 1971 Amendment, § 1(e), 85 Stat. 391 (formerly codified as 17 U. S. C. §26) (“For the purposes of [specified sections, not including § 1(a)], but not for any other purpose, a reproduc­tion of a [sound recording] shall be considered to be a copy thereof”). This concept is carried forward into the 1976 Act, which distinguishes between “copies” and “phonoreeords.” See n. 7, supra.

25

During consideration of the 1976 Act, Congress, of course, was well aware of the limited nature of the protection granted to sound recordings under the 1971 Amendment. See 1975 House Hearings 113 (testimony of Barbara Ringer, Register of Copyrights) (1971 Amendment “created a copyright in a sound recording . . . but limited it to the particular situation of so-called piracy”); id., at 1380 (letter from John Lorenz, Acting Librar­ian of Congress) (under 1971 Amendment “only the unauthorized reproduc­tion and distribution to the public of copies of the sound recording is pro­hibited. Thus, the duplication of sound recordings for private, personal use and the performance of sound recordings through broadcasting or other means are outside the scope of the amendment”).

26

Representative Kastenmeier, the principal House sponsor of the 1976 revision bill and Chairman of the House Subcommittee that produced it, made this explicit on the opening day of the House hearings:

“[F]rom time to time, certain areas have not been covered in the bill. But is it not the case, this being a unified code, that the operation of the bill does apply whether or not we specifically deal with a subject or not? . . .

“Therefore, we can really not fail to deal with an issue. It will be dealt with one way or the other. The code, title 17, will cover it. So we have made a conscientious decision even by omission. . . .

“. . . By virtue of passing this bill, we will deal with every issue. Whether we deal with it completely or not for the purpose of resolving the issues involved is the only question, not whether it has dealt with the four corners of the bill because the four corners of the bill will presume to deal with everything in copyright.” Id., at 115.

27

The precise phrase “fair use” apparently did not enter the case law until 1869, see Lawrence v. Dana, 15 F. Cas. 26, 60 (No. 8,136) (CC Mass.), but the doctrine itself found early expression in Folsom v. Marsh, 9 F. Cas. 342 (No. 4,901) (CC Mass. 1841). Justice Story was faced there with the “intricate and embarrassing questio[n]” whether a biography con­taining copyrighted letters was “a justifiable use of the original materials, such as the law recognizes as no infringement of the copyright of the plain­tiffs.” Id., at 344, 348. In determining whether the use was permitted, it was necessary, said Justice Story, to consider “the nature and objects of the selections made, the quantity and value of the materials used, and the degree in which the use may prejudice the sale, or diminish the profits, or supersede the objects, of the original work. . . . Much must, in such cases, depend upon the nature of the new work, the value and extent of the copies, and the degree in which the original authors may be injured thereby.” Id., at 348-349.

Similar lists were compiled by later courts. See, e. g., Tennessee Fabri­cating Co. v. Moultrie Mfg. Co., 421 F. 2d 279, 283 (CA5), cert. denied, 398 U. S. 928 (1970); Mathews Conveyer Co. v. Palmer-Bee Co., 135 F. 2d 73, 85 (CA6 1943); Columbia Pictures Corp. v. National Broadcasting Co., 137 F. Supp. 348 (SD Cal. 1955); Shapiro, Bernstein & Co. v. P. F. Collier & Son Co., 26 USPQ 40, 43 (SDNY 1934); Hill v. Whalen & Martell, Inc., 220 F. 359, 360 (SDNY 1914).

28

“The world goes ahead because each of us builds on the work of our predecessors. ‘A dwarf standing on the shoulders of a giant can see far­ther than the giant himself.’” Chafee, Reflections on the Law of Copy­right: I, 45 Colum. L. Rev. 503, 511 (1945).

29

Quoting from the Register’s 1961 Report, the Senate and House Re­ports give examples of possible fair uses:

“ ‘quotation of excerpts in a review or criticism for purposes of illustration or comment; quotation of short passages in a scholarly or technical work, for illustration or clarification of the author’s observations; use in a parody of some of the content of the work parodied; summary of an address or arti­cle, with brief quotations, in a news report; reproduction by a library of a portion of a work to replace part of a damaged copy; reproduction by a teacher or student of a small part of a work to illustrate a lesson; reproduc­tion of a work in legislative or judicial proceedings or reports; incidental and fortuitous reproduction, in a newsreel or broadcast, of a work located in the scene of an event being reported.’ ” 1975 Senate Report 61-62; 1976 House Report 65.

30

See, e. g., Triangle Publications, Inc. v. Knight-Ridder Newspapers, Inc., 626 F. 2d 1171 (CA5 1980) (comparative advertising).

31

Professor Seltzer has characterized these lists of uses as “reflectfing] what in fact the subject matter of fair use has in the history of its adjudica­tion consisted in: it has always had to do with the use by a second author of a first author’s work.” L. Seltzer, Exemptions and Fair Use in Copyright 24 (1978) (emphasis removed). He distinguishes “the mere reproduction of a work in order to use it for its intrinsic purpose — to make what might be called the ‘ordinary’ use of it.” When copies are made for “ordinary” use of the work, “ordinary infringement has customarily been triggered, not notions of fair use” (emphasis in original). Ibid. See also 3 M. Nimmer, Copyright § 13.05[A][1] (1983) (“Use of a work in each of the foregoing con­texts either necessarily or usually involves its use in a derivative work”).

32

Williams & Wilkins Co. v. United States, 203 Ct. Cl. 74, 487 F. 2d 1345 (1973), aff’d by an equally divided Court, 420 U. S. 376 (1975), in­volved the photocopying of scientific journal articles; the Court of Claims stressed that the libraries performing the copying were “devoted solely to the advancement and dissemination of medical knowledge,” 203 Ct. Cl., at 91, 487 F. 2d, at 1354, and that “medical science would be seriously hurt if such library photocopying were stopped.” Id., at 95, 487 F. 2d, at 1356.

The issue of library copying is now covered by § 108 of the 1976 Act. That section, which Congress regarded as “authorizing] certain photo­copying practices which may not qualify as a fair use,” 1975 Senate Report 67; 1976 House Report 74, permits the making of copies only for “private study, scholarship, or research.” §§ 108(d)(1) and (e)(1).

33

In the words of Lord Mansfield: “[W]e must take care to guard against two extremes equally prejudicial; the one, that men of ability, who have employed their time for the service of the community, may not be deprived of their just merits, and the reward of their ingenuity and labour; the other, that the world may not be deprived of improvements, nor the progress of the arts be retarded.” Sayre v. Moore, as set forth in Cary v. Longman, 1 East 358, 361, n. (b), 102 Eng. Rep. 138, 140, n. (b) (K. B. 1785). See Register’s Supplementary Report 13.

34

This point was brought home repeatedly by the Register of Copy­rights. Mentioning the “multitude of technological developments” since passage of the 1909 Act, including “remarkable developments in the use of video tape,” Register’s Supplementary Report xiv-xv, the Register cautioned:

“I realize, more clearly now than I did in 1961, that the revolution in communications has brought with it a serious challenge to the author’s copyright. This challenge comes not only from the ever-growing commer­cial interests who wish to use the author’s works for private gain. An equally serious attack has come from people with a sincere interest in the public welfare who fully recognize . . . ‘that the real heart of civiliza­tion . . . owes its existence to the author’; ironically, in seeking to make the author’s works widely available by freeing them from copyright re­strictions, they fail to realize that they are whittling away the very thing that nurtures authorship in the first place. An accommodation among con­flicting demands must be worked out, true enough, but not by denying the fundamental constitutional directive: to encourage cultural progress by securing the author’s exclusive rights to him for a limited time.” Id., at xv; see 1975 House Hearings 117 (testimony of Barbara Ringer, Regis­ter of Copyrights).

35

A VTR owner who has taped a favorite movie for repeated viewing will be less likely to rent or buy a tape containing the same movie, watch a tele­vised rerun, or pay to see the movie at a theater. Although time-shifting may not replace theater or rerun viewing or the purchase of prerecorded tapes or discs, it may well replace rental usage; a VTR user who has re­corded a first-run movie for later viewing will have no need to rent a copy when he wants to see it. Both library-builders and time-shifters may avoid commercials; the library-builder may use the pause control to record without them, and all users may fast-forward through commercials on playback.

The Studios introduced expert testimony that both time-shifting and librarying would tend to decrease their revenue from copyrighted works. See 480 F. Supp., at 440. The District Court’s findings also show substan­tial library-building and avoidance of commercials. Both sides submitted surveys showing that the average Betamax user owns between 25 and 32 tapes. The Studios’ survey showed that at least 40% of users had more than 10 tapes in a “library”; Sony’s survey showed that more than 40% of users planned to view their tapes more than once; and both sides’ surveys showed that commercials were avoided at least 25% of the time. Id., at 438-439.

36

Concern over the impact of a use upon “potential” markets is to be found in cases decided both before and after § 107 lent Congress’ imprima­tur to the judicially created doctrine of fair use. See, e. g., Iowa State University Research Foundation, Inc. v. American Broadcasting Cos., 621 F. 2d 57, 60 (CA2 1980) (“the effect of the use on the copyright holder’s potential market for the work”); Meeropol v. Nizer, 560 F. 2d 1061, 1070 (CA2 1977) (“A key issue in fair use cases is whether the defendant’s work tends to diminish or prejudice the potential sale of plaintiff’s work”), cert. denied, 434 U. S. 1013 (1978); Williams & Wilkins Co. v. United States, 203 Ct. Cl., at 88, 487 F. 2d, at 1352 (“the effect of the use on a copyright owner’s potential market for and value of his work”); Encyclopaedia Bri­tannica Educational Corp. v. Crooks, 542 F. Supp. 1156, 1173 (WDNY 1982) (“[T]he concern here must be focused on a copyrighted work’s poten­tial market. It is perfectly possible that plaintiffs’ profits would have been greater, but for the kind of videotaping in question”) (emphasis in original).

37

This intent is manifested further by provisions of the 1976 Act that exempt from liability persons who, while not participating directly in any infringing activity, could otherwise be charged with contributory infringe­ment. See § 108(f)(1) (library not liable “for the unsupervised use of re­producing equipment located on its premises,” provided that certain warn­ings are posted); § 110(6) (“governmental body” or “nonprofit agricultural or horticultural organization” not liable for infringing performance by con­cessionaire “in the course of an annual agricultural or horticultural fair or exhibition”).

38

In Screen Gems, on which the Gershwin court relied, the court held that liability could be imposed on a shipper of unauthorized “bootleg” records and a radio station that broadcast advertisements of the records, provided they knew or should have known that the records were infring­ing. The court concluded that the records’ low price and the manner in which the records were marketed could support a finding of “constructive knowledge” even if actual knowledge were not shown.

39

See, e. g., Famous Music Corp. v. Bay State Harness Horse Racing & Breeding Assn., Inc., 554 F. 2d 1213 (CA1 1977); Dreamland Ball Room, Inc. v. Shapiro, Bernstein & Co., 36 F. 2d 354 (CA7 1929); M. Witmark & Sons v. Tremont Social & Athletic Club, 188 F. Supp. 787, 790 (Mass. 1960); see also Twentieth Century Music Corp. v. Aiken, 422 U. S. 151, 157 (1975); Buck v. Jewell-LaSalle Realty Co., 283 U. S. 191, 198-199 (1931); 3 M. Nimmer, Copyright § 12.04[A], p. 12-35 (1983).

Courts have premised liability in these cases on the notion that the de­fendant had the ability to supervise or control the infringing activities, see, e. g., Shapiro, Bernstein & Co. v. H. L. Green Co., 316 F. 2d 304, 307 (CA2 1963); KECA Music, Inc. v. Dingus McGee’s Co., 432 F. Supp. 72, 74 (WD Mo. 1977). This notion, however, is to some extent fictional; the defendant cannot escape liability by instructing the performers not to play copyrighted music, or even by inserting a provision to that effect into the performers’ contract. Famous Music Corp. v. Bay State Harness Horse Racing & Breeding Assn., Inc., 554 F. 2d, at 1214-1215; KECA Music, Inc. v. Dingus McGee’s Co., 432 F. Supp., at 75; Shapiro, Bernstein & Co. v. Veltin, 47 F. Supp. 648, 649 (WD La. 1942). Congress expressly re­jected a proposal to exempt proprietors from this type of liability under the 1976 Act. See 1975 Senate Report 141-142; 1976 House Report 159-160; 1975 House Hearings 1812-1813 (testimony of Barbara Ringer, Register of Copyrights); id., at 1813 (colloquy between Rep. Pattison and Barbara Ringer).

The Court’s attempt to distinguish these cases on the ground of “con­trol,” ante, at 437, is obviously unpersuasive. The direct infringer ordi­narily is not employed by the person held liable; instead, he is an independ­ent contractor. Neither is he always an agent of the person held liable; Screen Gems makes this apparent.

40

My conclusion respecting contributory infringement does not include the retailer defendants. The District Court found that one of the retailer defendants had assisted in the advertising campaign for the Betamax, but made no other findings respecting their knowledge of the Betamax’s in­tended uses. I do not agree with the Court of Appeals, at least on this record, that the retailers “are sufficiently engaged in the enterprise to be held accountable,” 659 F. 2d 963, 976 (1981). In contrast, the advertising agency employed to promote the Betamax was far more actively engaged in the advertising campaign, and petitioners have not argued that the agency’s liability differs in any way from that of Sony Corporation and Sony Corporation of America.

41

The “staple article of commerce” doctrine protects those who manufac­ture products incorporated into or used with patented inventions — for ex­ample, the paper and ink used with patented printing machines, Henry v. A. B. Dick Co., 224 U. S. 1 (1912), or the dry ice used with patented refrig­eration systems, Carbice Corp. v. American Patents Corp., 283 U. S. 27 (1931). Because a patent holder has the right to control the use of the patented item as well as its manufacture, see Motion Picture Patents Co. v. Universal Film Mfg. Co., 243 U. S. 502, 509-510 (1917); 35 U. S. C. § 271(a), such protection for the manufacturer of the incorporated product is necessary to prevent patent holders from extending their monopolies by suppressing competition in unpatented components and supplies suitable for use with the patented item. See Dawson Chemical Co. v. Rohm & Haas Co., 448 U. S. 176, 197-198 (1980). The doctrine of contributory patent infringement has been the subject of attention by the courts and by Congress, see id., at 202-212, and has been codified since 1952, 66 Stat. 792, but was never mentioned during the copyright law revision process as having any relevance to contributory copyright infringement.

42

Although VTR’s also may be used to watch prerecorded video cassettes and to make home motion pictures, these uses do not require a tuner such as the Betamax contains. See n. 1, supra. The Studios do not object to Sony’s sale of VTR’s without tuners. Brief for Respondents 5, n. 9. In considering the noninfringing uses of the Betamax, therefore, those uses that would remain possible without the Betamax’s built-in tuner should not be taken into account.

43

Noninfringing uses would include, for example, recording works that are not protected by copyright, recording works that have entered the public domain, recording with permission of the copyright owner, and, of course, any recording that qualifies as fair use. See, e. g., Bruzzone v. Miller Brewing Co., 202 USPQ 809 (ND Cal. 1979) (use of home VTR for market research studies).

44

Sony asserts that much or most television broadcasting is available for home recording because (1) no copyright owner other than the Studios has brought an infringement action, and (2) much televised material is ineligi­ble for copyright protection because videotapes of the broadcasts are not kept. The first of these assertions is irrelevant; Sony’s liability does not turn on the fact that only two copyright owners thus far have brought suit. The amount of infringing use must be determined through consideration of the television market as a whole. Sony’s second assertion is based on a faulty premise; the Copyright Office permits audiovisual works transmit­ted by television to be registered by deposit of sample frames plus a de­scription of the work. See 37 CFR §§ 202.20(c)(2)(ii) and 202.21(g) (1983). Moreover, although an infringement action cannot be brought unless the work is registered, 17 U. S. C. § 411(a) (1982 ed.), registration is not a con­dition of copyright protection. § 408(a). Copying an unregistered work still may be infringement. Cf. § 506(a) (liability for criminal copyright infringement; not conditioned on prior registration).

45

Even if concern with remedy were appropriate at the. liability stage, the Court's use of the District Court’s findings is’ somewhat cavalier. The Court relies heavily on testimony by representatives of professional sports leagues to the effect that they have no objection to VTR recording. The Court never states, however, whether the sports leagues are copyright holders, and if so, whether they have exclusive copyrights to sports broad­casts. It is therefore unclear whether the sports leagues have authority to consent to copying the broadcasts of their events.

Assuming that the various sports leagues do have exclusive copyrights in some of their broadcasts, the amount of authorized time-shifting still would not be overwhelming. Sony’s own survey indicated that only 7.3% of all Betamax use is to record sports events of all kinds. Tr. 2353, De­fendants’ Exh. OT, Table 20. Because Sony’s witnesses did not represent all forms of sports events, moreover, this figure provides only a tenuous basis for this Court to engage in factfinding of its own.

The only witness at trial who was clearly an exclusive copyright owner and who expressed no objection to unauthorized time-shifting was the owner of the copyright in Mister Rogers’ Neighborhood. But the Court cites no evidence in the record to the effect that anyone makes VTR copies of that program. The simple fact is that the District Court made no find­ings on the amount of authorized time-shifting that takes place. The Court seems to recognize this gap in its reasoning, and phrases its argu­ment as a hypothetical. The Court states: “If there are millions of owners of VTR’s who make copies of televised sports events, religious broadcasts, and educational programs such as Mister Rogers’ Neighborhood, and 2/the proprietors of those programs welcome the practice,” the sale of VTR’s “should not be stifled” in order to protect respondents’ copyrights. Ante, at 446 (emphasis supplied). Given that the Court seems to recognize that its argument depends on findings that have not been made, it seems that a remand is inescapable.

46

As has been explained, some uses of time-shifting, such as copying an old newspaper clipping for a friend, are fair use because of their de minimis effect on the copyright holder. The scale of copying involved in this case, of course, is of an entirely different magnitude, precluding appli­cation of such an exception.

47

Home Recording of Copyrighted Works: Hearing before the Sub­committee on Courts, Civil Liberties, and the Administration of Justice of the House Committee on the Judiciary, 97th Cong., 2d Sess., pt. 2, p. 1260 (1982) (memorandum of Prof. Laurence H. Tribe).

48

See A Survey of Betamax Owners, Tr. 2353, Defendants’ Exh. OT, Table 20, cited in Brief for Respondents 52.

49

The Court’s one oblique acknowledgment of this third factor, ante, at 447, and n. 30, seems to suggest that the fact that time-shifting involves copying complete works is not very significant because the viewers already have been asked to watch the initial broadcast free. This suggestion misses the point. As has been noted, a book borrowed from a public li­brary may not be copied any more freely than one that has been purchased. An invitation to view a showing is completely different from an invitation to copy a copyrighted work.

50

The Court implicitly has recognized that this market is very significant. The central concern underlying the Court's entire opinion is that there is a large audience who would like very much to be able to view programs at times other than when they are broadcast. Ante, at 446. The Court sim­ply misses the implication of its own concerns.

51

Other nations have imposed royalties on the manufacturers of products used to infringe copyright. See, e. g., Copyright Laws and Treaties of the World (UNESCO/BNA 1982) (English translation), reprinting Federal Act on Copyright in Works of Literature and Art and on Related Rights (Aus­tria), §§ 42(5)-(7), and An Act dealing with Copyright and Related Rights (Federal Republic of Germany), Art. 53(5). A study produced for the Commission of European Communities has recommended that these re­quirements “serve as a pattern” for the European community. A. Dietz, Copyright Law in the European Community 135 (1978). While these roy­alty systems ordinarily depend on the existence of authors’ collecting soci­eties, see id., at 119, 136, such collecting societies are a familiar part of our copyright law. See generally Broadcast Music, Inc. v. Columbia Broad­casting System, Inc., 441 U. S. 1, 4-5 (1979). Fashioning relief of this sort, of course, might require bringing other copyright owners into court through certification of a class or otherwise.

Respondent companies distribute free software that allows computer users to share electronic files through peer-to-peer networks, so called because the computers communicate directly with each other, not through central servers. Although such networks can be used to share any type of digital file, recipients of respondents' software have mostly used them to share copyrighted music and video files without authorization. Seeking damages and an injunction, a group of movie studios and other copyright holders (hereinafter MGM) sued respondents for their users' copyright infringements, alleging that respondents knowingly and intentionally distributed their software to enable users to infringe copyrighted works in violation of the Copyright Act.

Discovery revealed that billions of files are shared across peer-to-peer networks each month. Respondents are aware that users employ their software primarily to download copyrighted files, although the decentralized networks do not reveal which files are copied, and when. Respondents have sometimes learned about the infringement directly when users have e-mailed questions regarding copyrighted works, and respondents have replied with guidance. Respondents are not merely passive recipients of information about infringement. The record is replete with evidence that when they began to distribute their free software, each of them clearly voiced the objective that recipients use the software to download copyrighted works and took active steps to encourage infringement. After the notorious file-sharing service, Napster, was sued by copyright holders for facilitating copyright infringement, both respondents promoted and marketed themselves as Napster alternatives. They receive no revenue from users, but, instead, generate income by selling advertising space, then streaming the advertising to their users. As the number of users increases, advertising opportunities are worth more. There is no evidence that either respondent made an effort to filter copyrighted material from users' downloads or otherwise to impede the sharing of copyrighted files.

While acknowledging that respondents' users had directly infringed MGM's copyrights, the District Court nonetheless granted respondents summary judgment as to liability arising from distribution of their software. [545 U.S. 914] The Ninth Circuit affirmed. It read Sony Corp. of America v. Universal City Studios, Inc., 464 U. S. 417, as holding that the distribution of a commercial product capable of substantial noninfringing uses could not give rise to contributory liability for infringement unless the distributor had actual knowledge of specific instances of infringement and failed to act on that knowledge. Because the appeals court found respondents' software to be capable of substantial noninfringing uses and because respondents had no actual knowledge of infringement owing to the software's decentralized architecture, the court held that they were not liable. It also held that they did not materially contribute to their users' infringement because the users themselves searched for, retrieved, and stored the infringing files, with no involvement by respondents beyond providing the software in the first place. Finally, the court held that respondents could not be held liable under a vicarious infringement theory because they did not monitor or control the software's use, had no agreed-upon right or current ability to supervise its use, and had no independent duty to police infringement.

Held: One who distributes a device with the object of promoting its use to infringe copyright, as shown by clear expression or other affirmative steps taken to foster infringement, going beyond mere distribution with knowledge of third-party action, is liable for the resulting acts of infringement by third parties using the device, regardless of the device's lawful uses. Pp. 928-941.

(a) The tension between the competing values of supporting creativity through copyright protection and promoting technological innovation by limiting infringement liability is the subject of this case. Despite offsetting considerations, the argument for imposing indirect liability here is powerful, given the number of infringing downloads that occur daily using respondents' software. When a widely shared product is used to commit infringement, it may be impossible to enforce rights in the protected work effectively against all direct infringers, so that the only practical alternative is to go against the device's distributor for secondary liability on a theory of contributory or vicarious infringement. One infringes contributorily by intentionally inducing or encouraging direct infringement, and infringes vicariously by profiting from direct infringement while declining to exercise the right to stop or limit it. Although "[t]he Copyright Act does not expressly render anyone liable for [another's] infringement," Sony, 464 U. S., at 434, these secondary liability doctrines emerged from common law principles and are well established in the law, e. g., id., at 486. Pp. 928-931.

(b) Sony addressed a claim that secondary liability for infringement can arise from the very distribution of a commercial product. There, [545 U.S. 915] copyright holders sued Sony, the manufacturer of videocassette recorders, claiming that it was contributorily liable for the infringement that occurred when VCR owners taped copyrighted programs. The evidence showed that the VCR's principal use was "time-shifting," i. e., taping a program for later viewing at a more convenient time, which the Court found to be a fair, noninfringing use. 464 U. S., at 423-424. Moreover, there was no evidence that Sony had desired to bring about taping in violation of copyright or taken active steps to increase its profits from unlawful taping. Id., at 438. On those facts, the only conceivable basis for liability was on a theory of contributory infringement through distribution of a product. Id., at 439. Because the VCR was "capable of commercially significant noninfringing uses," the Court held that Sony was not liable. Id., at 442. This theory reflected patent law's traditional staple article of commerce doctrine that distribution of a component of a patented device will not violate the patent if it is suitable for use in other ways. 35 U. S. C. § 271(c). The doctrine absolves the equivocal conduct of selling an item with lawful and unlawful uses and limits liability to instances of more acute fault. In this case, the Ninth Circuit misread Sony to mean that when a product is capable of substantial lawful use, the producer cannot be held contributorily liable for third parties' infringing use of it, even when an actual purpose to cause infringing use is shown, unless the distributors had specific knowledge of infringement at a time when they contributed to the infringement and failed to act upon that information. Sony did not displace other secondary liability theories. Pp. 931-934.

(c) Nothing in Sony requires courts to ignore evidence of intent to promote infringement if such evidence exists. It was never meant to foreclose rules of fault-based liability derived from the common law. 464 U. S., at 439. Where evidence goes beyond a product's characteristics or the knowledge that it may be put to infringing uses, and shows statements or actions directed to promoting infringement, Sony's staple-article rule will not preclude liability. At common law a copyright or patent defendant who "not only expected but invoked [infringing use] by advertisement" was liable for infringement. Kalem Co. v. Harper Brothers, 222 U. S. 55, 62-63. The rule on inducement of infringement as developed in the early cases is no different today. Evidence of active steps taken to encourage direct infringement, such as advertising an infringing use or instructing how to engage in an infringing use, shows an affirmative intent that the product be used to infringe, and overcomes the law's reluctance to find liability when a defendant merely sells a commercial product suitable for some lawful use. A rule that premises liability on purposeful, culpable expression and conduct [545 U.S. 916] does nothing to compromise legitimate commerce or discourage innovation having a lawful promise. Pp. 934-937.

(d) On the record presented, respondents' unlawful objective is unmistakable. The classic instance of inducement is by advertisement or solicitation that broadcasts a message designed to stimulate others to commit violations. MGM argues persuasively that such a message is shown here. Three features of the evidence of intent are particularly notable. First, each of the respondents showed itself to be aiming to satisfy a known source of demand for copyright infringement, the market comprising former Napster users. Respondents' efforts to supply services to former Napster users indicate a principal, if not exclusive, intent to bring about infringement. Second, neither respondent attempted to develop filtering tools or other mechanisms to diminish the infringing activity using their software. While the Ninth Circuit treated that failure as irrelevant because respondents lacked an independent duty to monitor their users' activity, this evidence underscores their intentional facilitation of their users' infringement. Third, respondents make money by selling advertising space, then by directing ads to the screens of computers employing their software. The more their software is used, the more ads are sent out and the greater the advertising revenue. Since the extent of the software's use determines the gain to the distributors, the commercial sense of their enterprise turns on high-volume use, which the record shows is infringing. This evidence alone would not justify an inference of unlawful intent, but its import is clear in the entire record's context. Pp. 937-940.

(e) In addition to intent to bring about infringement and distribution of a device suitable for infringing use, the inducement theory requires evidence of actual infringement by recipients of the device, the software in this case. There is evidence of such infringement on a gigantic scale. Because substantial evidence supports MGM on all elements, summary judgment for respondents was error. On remand, reconsideration of MGM's summary judgment motion will be in order. Pp. 940-941.

380 F. 3d 1154, vacated and remanded.

SOUTER, J., delivered the opinion for a unanimous Court. GINSBURG, J., filed a concurring opinion, in which REHNQUIST, C. J., and KENNEDY, J., joined, post, p. 942. BREYER, J., filed a concurring opinion, in which STEVENS and O'CONNOR, JJ., joined, post, p. 949.

CERTIORARI TO THE UNITED STATES COURT OF APPEALS FOR THE NINTH CIRCUIT

Donald B. Verrilli, Jr., argued the cause for petitioners. With him on the briefs for the motion picture studio and recording company petitioners were Ian Heath Gershengorn, [545 U.S. 917] William M. Hohengarten, Steven B. Fabrizio, Thomas J. Perrelli, Matthew J. Oppenheim, David E. Kendall, Thomas G. Hentoff, Kenneth W. Starr, Russell J. Frackman, George M. Borkowski, Robert M. Schwartz, Gregory P. Goeckner, Dean C. Garfield, Elaine J. Goldenberg, Matthew Hersh, Steven M. Marks, and Stanley Pierre-Louis. Carey R. Ramos, Peter L. Felcher, Aidan Synnott, Theodore K. Cheng, Kelli L. Sager, Andrew J. Thomas, Jeffrey H. Blum, and Jeffrey L. Fisher filed briefs for the songwriter and music publisher petitioners.

Acting Solicitor General Clement argued the cause for the United States as amicus curiae urging reversal. With him on the brief were Assistant Attorney General Keisler, Deputy Solicitor General Hungar, Douglas H. Hallward-Driemeier, Anthony A. Yang, David O. Carson, and John M. Whealan.

Richard G. Taranto argued the cause for respondents. With him on the brief were H. Bartow Farr III, Cindy A. Cohn, Fred Von Lohmann, Michael H. Page, Mark A. Lemley, Charles S. Baker, and Matthew A. Neco.^[*]

[545 U.S. 918] JUSTICE SOUTER delivered the opinion of the Court.

The question is under what circumstances the distributor of a product capable of both lawful and unlawful use is liable [545 U.S. 919] for acts of copyright infringement by third parties using the product. We hold that one who distributes a device with the object of promoting its use to infringe copyright, as shown by clear expression or other affirmative steps taken to foster infringement, is liable for the resulting acts of infringement by third parties.

Respondents, Grokster, Ltd., and StreamCast Networks, Inc., defendants in the trial court, distribute free software products that allow computer users to share electronic files through peer-to-peer networks, so called because users' computers communicate directly with each other, not through [545 U.S. 920] central servers. The advantage of peer-to-peer networks over information networks of other types shows up in their substantial and growing popularity. Because they need no central computer server to mediate the exchange of information or files among users, the high-bandwidth communications capacity for a server may be dispensed with, and the need for costly server storage space is eliminated. Since copies of a file (particularly a popular one) are available on many users' computers, file requests and retrievals may be faster than on other types of networks, and since file exchanges do not travel through a server, communications can take place between any computers that remain connected to the network without risk that a glitch in the server will disable the network in its entirety. Given these benefits in security, cost, and efficiency, peer-to-peer networks are employed to store and distribute electronic files by universities, government agencies, corporations, and libraries, among others.^[1]

Other users of peer-to-peer networks include individual recipients of Grokster's and StreamCast's software, and although the networks that they enjoy through using the software can be used to share any type of digital file, they have prominently employed those networks in sharing copyrighted music and video files without authorization. A group of copyright holders (MGM for short, but including motion picture studios, recording companies, songwriters, and music publishers) sued Grokster and StreamCast for their users' copyright infringements, alleging that they [545 U.S. 921] knowingly and intentionally distributed their software to enable users to reproduce and distribute the copyrighted works in violation of the Copyright Act, 17 U. S. C. § 101 et seq. (2000 ed. and Supp. II).^[2] MGM sought damages and an injunction.

Discovery during the litigation revealed the way the software worked, the business aims of each defendant company, and the predilections of the users. Grokster's eponymous software employs what is known as FastTrack technology, a protocol developed by others and licensed to Grokster. StreamCast distributes a very similar product except that its software, called Morpheus, relies on what is known as Gnutella technology.^[3] A user who downloads and installs either software possesses the protocol to send requests for files directly to the computers of others using software compatible with FastTrack or Gnutella. On the FastTrack network opened by the Grokster software, the user's request goes to a computer given an indexing capacity by the software and designated a supernode, or to some other computer with comparable power and capacity to collect temporary indexes of the files available on the computers of users connected to it. The supernode (or indexing computer) searches its own index and may communicate the search request to other supernodes. If the file is found, the supernode discloses its location to the computer requesting it, and the requesting user can download the file directly from the computer located. The copied file is placed in a designated sharing folder on the requesting user's computer, where it is available for other users to download in turn, along with any other file in that folder.[545 U.S. 922] In the Gnutella network made available by Morpheus, the process is mostly the same, except that in some versions of the Gnutella protocol there are no supernodes. In these versions, peer computers using the protocol communicate directly with each other. When a user enters a search request into the Morpheus software, it sends the request to computers connected with it, which in turn pass the request along to other connected peers. The search results are communicated to the requesting computer, and the user can download desired files directly from peers' computers. As this description indicates, Grokster and StreamCast use no servers to intercept the content of the search requests or to mediate the file transfers conducted by users of the software, there being no central point through which the substance of the communications passes in either direction.^[4]

Although Grokster and StreamCast do not therefore know when particular files are copied, a few searches using their software would show what is available on the networks the software reaches. MGM commissioned a statistician to conduct a systematic search, and his study showed that nearly 90% of the files available for download on the FastTrack system were copyrighted works.^[5] Grokster and StreamCast dispute this figure, raising methodological problems and arguing that free copying even of copyrighted works may be authorized by the rightholders. They also argue that potential noninfringing uses of their software are significant in kind, even if infrequent in practice. Some musical performers, for example, have gained new audiences by distributing [545 U.S. 923] their copyrighted works for free across peer-to-peer networks, and some distributors of unprotected content have used peer-to-peer networks to disseminate files, Shakespeare being an example. Indeed, StreamCast has given Morpheus users the opportunity to download the briefs in this very case, though their popularity has not been quantified.

As for quantification, the parties' anecdotal and statistical evidence entered thus far to show the content available on the FastTrack and Gnutella networks does not say much about which files are actually downloaded by users, and no one can say how often the software is used to obtain copies of unprotected material. But MGM's evidence gives reason to think that the vast majority of users' downloads are acts of infringement, and because well over 100 million copies of the software in question are known to have been downloaded, and billions of files are shared across the FastTrack and Gnutella networks each month, the probable scope of copyright infringement is staggering.

Grokster and StreamCast concede the infringement in most downloads, Brief for Respondents 10, n. 6, and it is uncontested that they are aware that users employ their software primarily to download copyrighted files, even if the decentralized FastTrack and Gnutella networks fail to reveal which files are being copied, and when. From time to time, moreover, the companies have learned about their users' infringement directly, as from users who have sent e-mail to each company with questions about playing copyrighted movies they had downloaded, to whom the companies have responded with guidance.^[6] App. 559-563, 808-816, 939-954. And MGM notified the companies of 8 million copyrighted files that could be obtained using their software.

Grokster and StreamCast are not, however, merely passive recipients of information about infringing use. The record is replete with evidence that from the moment Grokster [545 U.S. 924] and StreamCast began to distribute their free software, each one clearly voiced the objective that recipients use it to download copyrighted works, and each took active steps to encourage infringement.

After the notorious file-sharing service, Napster, was sued by copyright holders for facilitation of copyright infringement, A&M Records, Inc. v. Napster, Inc., 114 F. Supp. 2d 896 (ND Cal. 2000), aff'd in part, rev'd in part, 239 F. 3d 1004 (CA9 2001), StreamCast gave away a software program of a kind known as OpenNap, designed as compatible with the Napster program and open to Napster users for downloading files from other Napster and OpenNap users' computers. Evidence indicates that "[i]t was always [StreamCast's] intent to use [its OpenNap network] to be able to capture email addresses of [its] initial target market so that [it] could promote [its] StreamCast Morpheus interface to them," App. 861; indeed, the OpenNap program was engineered "`to leverage Napster's 50 million user base,'" id., at 746.

StreamCast monitored both the number of users downloading its OpenNap program and the number of music files they downloaded. Id., at 859, 863, 866. It also used the resulting OpenNap network to distribute copies of the Morpheus software and to encourage users to adopt it. Id., at 861, 867, 1039. Internal company documents indicate that StreamCast hoped to attract large numbers of former Napster users if that company was shut down by court order or otherwise, and that StreamCast planned to be the next Napster. Id., at 861. A kit developed by StreamCast to be delivered to advertisers, for example, contained press articles about StreamCast's potential to capture former Napster users, id., at 568-572, and it introduced itself to some potential advertisers as a company "which is similar to what Napster was," id., at 884. It broadcast banner advertisements to users of other Napster-compatible software, urging them to adopt its OpenNap. Id., at 586. An internal e-mail from a company executive stated: "`We have put this network in [545 U.S. 925] place so that when Napster pulls the plug on their free service . . . or if the Court orders them shut down prior to that . . . we will be positioned to capture the flood of their 32 million users that will be actively looking for an alternative.'" Id., at 588-589, 861.

Thus, StreamCast developed promotional materials to market its service as the best Napster alternative. One proposed advertisement read: "Napster Inc. has announced that it will soon begin charging you a fee. That's if the courts don't order it shut down first. What will you do to get around it?" Id., at 897. Another proposed ad touted StreamCast's software as the "#1 alternative to Napster" and asked "[w]hen the lights went off at Napster . . . where did the users go?" Id., at 836 (ellipsis in original).^[7] StreamCast even planned to flaunt the illegal uses of its software; when it launched the OpenNap network, the chief technology officer of the company averred that "[t]he goal is to get in trouble with the law and get sued. It's the best way to get in the new[s]." Id., at 916.

The evidence that Grokster sought to capture the market of former Napster users is sparser but revealing, for Grokster launched its own OpenNap system called Swaptor and inserted digital codes into its Web site so that computer users using Web search engines to look for "Napster" or "[f]ree file sharing" would be directed to the Grokster Web site, where they could download the Grokster software. Id., at 992-993. And Grokster's name is an apparent derivative of Napster.

StreamCast's executives monitored the number of songs by certain commercial artists available on their networks, and an internal communication indicates they aimed to have a larger number of copyrighted songs available on their networks [545 U.S. 926] than other file-sharing networks. Id., at 868. The point, of course, would be to attract users of a mind to infringe, just as it would be with their promotional materials developed showing copyrighted songs as examples of the kinds of files available through Morpheus. Id., at 848. Morpheus in fact allowed users to search specifically for "Top 40" songs, id., at 735, which were inevitably copyrighted. Similarly, Grokster sent users a newsletter promoting its ability to provide particular, popular copyrighted materials. Brief for Motion Picture Studio and Recording Company Petitioners 7-8.

In addition to this evidence of express promotion, marketing, and intent to promote further, the business models employed by Grokster and StreamCast confirm that their principal object was use of their software to download copyrighted works. Grokster and StreamCast receive no revenue from users, who obtain the software itself for nothing. Instead, both companies generate income by selling advertising space, and they stream the advertising to Grokster and Morpheus users while they are employing the programs. As the number of users of each program increases, advertising opportunities become worth more. Cf. App. 539, 804. While there is doubtless some demand for free Shakespeare, the evidence shows that substantive volume is a function of free access to copyrighted work. Users seeking Top 40 songs, for example, or the latest release by Modest Mouse, are certain to be far more numerous than those seeking a free Decameron, and Grokster and StreamCast translated that demand into dollars.

Finally, there is no evidence that either company made an effort to filter copyrighted material from users' downloads or otherwise impede the sharing of copyrighted files. Although Grokster appears to have sent e-mails warning users about infringing content when it received threatening notice from the copyright holders, it never blocked anyone from continuing to use its software to share copyrighted files. [545 U.S. 927] Id., at 75-76. StreamCast not only rejected another company's offer of help to monitor infringement, id., at 928-929, but blocked the Internet Protocol addresses of entities it believed were trying to engage in such monitoring on its networks, id., at 917-922.

After discovery, the parties on each side of the case crossmoved for summary judgment. The District Court limited its consideration to the asserted liability of Grokster and StreamCast for distributing the current versions of their software, leaving aside whether either was liable "for damages arising from past versions of their software, or from other past activities." 259 F. Supp. 2d 1029, 1033 (CD Cal. 2003). The District Court held that those who used the Grokster and Morpheus software to download copyrighted media files directly infringed MGM's copyrights, a conclusion not contested on appeal, but the court nonetheless granted summary judgment in favor of Grokster and StreamCast as to any liability arising from distribution of the then-current versions of their software. Distributing that software gave rise to no liability in the court's view, because its use did not provide the distributors with actual knowledge of specific acts of infringement. Case No. CV 01 08541 SVW (PJWx) (CD Cal., June 18, 2003), App. 1213.

The Court of Appeals affirmed. 380 F. 3d 1154 (CA9 2004). In the court's analysis, a defendant was liable as a contributory infringer when it had knowledge of direct infringement and materially contributed to the infringement. But the court read Sony Corp. of America v. Universal City Studios, Inc., 464 U. S. 417 (1984), as holding that distribution of a commercial product capable of substantial noninfringing uses could not give rise to contributory liability for infringement unless the distributor had actual knowledge of specific instances of infringement and failed to act on that knowledge. The fact that the software was capable of substantial noninfringing uses in the Ninth Circuit's view meant [545 U.S. 928] that Grokster and StreamCast were not liable, because they had no such actual knowledge, owing to the decentralized architecture of their software. The court also held that Grokster and StreamCast did not materially contribute to their users' infringement because it was the users themselves who searched for, retrieved, and stored the infringing files, with no involvement by the defendants beyond providing the software in the first place.

The Ninth Circuit also considered whether Grokster and StreamCast could be liable under a theory of vicarious infringement. The court held against liability because the defendants did not monitor or control the use of the software, had no agreed-upon right or current ability to supervise its use, and had no independent duty to police infringement. We granted certiorari. 543 U. S. 1032 (2004).

MGM and many of the amici fault the Court of Appeals's holding for upsetting a sound balance between the respective values of supporting creative pursuits through copyright protection and promoting innovation in new communication technologies by limiting the incidence of liability for copyright infringement. The more artistic protection is favored, the more technological innovation may be discouraged; the administration of copyright law is an exercise in managing the tradeoff. See Sony Corp. v. Universal City Studios, supra, at 442; see generally Ginsburg, Copyright and Control Over New Technologies of Dissemination, 101 Colum. L. Rev. 1613 (2001); Lichtman & Landes, Indirect Liability for Copyright Infringement: An Economic Perspective, 16 Harv. J. L. & Tech. 395 (2003).

The tension between the two values is the subject of this case, with its claim that digital distribution of copyrighted material threatens copyright holders as never before, because every copy is identical to the original, copying is easy, [545 U.S. 929] and many people (especially the young) use file-sharing software to download copyrighted works. This very breadth of the software's use may well draw the public directly into the debate over copyright policy, Peters, Brace Memorial Lecture: Copyright Enters the Public Domain, 51 J. Copyright Soc. 701, 705-717 (2004) (address by Register of Copyrights), and the indications are that the ease of copying songs or movies using software like Grokster's and Napster's is fostering disdain for copyright protection, Wu, When Code Isn't Law, 89 Va. L. Rev. 679, 724-726 (2003). As the case has been presented to us, these fears are said to be offset by the different concern that imposing liability, not only on infringers but on distributors of software based on its potential for unlawful use, could limit further development of beneficial technologies. See, e. g., Lemley & Reese, Reducing Digital Copyright Infringement Without Restricting Innovation, 56 Stan. L. Rev. 1345, 1386-1390 (2004); Brief for Innovation Scholars and Economists as Amici Curiae 15-20; Brief for Emerging Technology Companies as Amici Curiae 19-25; Brief for Intel Corporation as Amicus Curiae 20-22.^[8]

The argument for imposing indirect liability in this case is, however, a powerful one, given the number of infringing downloads that occur every day using StreamCast's and Grokster's software. When a widely shared service or product is used to commit infringement, it may be impossible to [545 U.S. 930] enforce rights in the protected work effectively against all direct infringers, the only practical alternative being to go against the distributor of the copying device for secondary liability on a theory of contributory or vicarious infringement. See In re Aimster Copyright Litigation, 334 F. 3d 643, 645-646 (CA7 2003).

One infringes contributorily by intentionally inducing or encouraging direct infringement, see Gershwin Pub. Corp. v. Columbia Artists Management, Inc., 443 F. 2d 1159, 1162 (CA2 1971), and infringes vicariously by profiting from direct infringement while declining to exercise a right to stop or limit it, Shapiro, Bernstein & Co. v. H. L. Green Co., 316 F. 2d 304, 307 (CA2 1963).^[9] Although "[t]he Copyright Act does not expressly render anyone liable for infringement committed by another," Sony Corp. v. Universal City Studios, 464 U. S., at 434, these doctrines of secondary liability emerged from common law principles and are well established in the law, id., at 486 (Blackmun, J., dissenting); Kalem Co. v. Harper Brothers, 222 U. S. 55, 62-63 (1911); Gershwin Pub. Corp. v. Columbia Artists Management, [545 U.S. 931] supra, at 1162; 3 M. Nimmer & D. Nimmer, Copyright § 12.04[A] (2005).

Despite the currency of these principles of secondary liability, this Court has dealt with secondary copyright infringement in only one recent case, and because MGM has tailored its principal claim to our opinion there, a look at our earlier holding is in order. In Sony Corp. v. Universal City Studios, supra, this Court addressed a claim that secondary liability for infringement can arise from the very distribution of a commercial product. There, the product, novel at the time, was what we know today as the videocassette recorder or VCR. Copyright holders sued Sony as the manufacturer, claiming it was contributorily liable for infringement that occurred when VCR owners taped copyrighted programs because it supplied the means used to infringe, and it had constructive knowledge that infringement would occur. At the trial on the merits, the evidence showed that the principal use of the VCR was for "`time-shifting,'" or taping a program for later viewing at a more convenient time, which the Court found to be a fair, not an infringing, use. Id., at 423-424. There was no evidence that Sony had expressed an object of bringing about taping in violation of copyright or had taken active steps to increase its profits from unlawful taping. Id., at 438. Although Sony's advertisements urged consumers to buy the VCR to "`record favorite shows'" or "`build a library'" of recorded programs, id., at 459 (Blackmun, J., dissenting), neither of these uses was necessarily infringing, id., at 424, 454-455.

On those facts, with no evidence of stated or indicated intent to promote infringing uses, the only conceivable basis for imposing liability was on a theory of contributory infringement arising from its sale of VCRs to consumers with knowledge that some would use them to infringe. Id., at 439. But because the VCR was "capable of commercially significant noninfringing uses," we held the manufacturer [545 U.S. 932] could not be faulted solely on the basis of its distribution. Id., at 442.

This analysis reflected patent law's traditional staple article of commerce doctrine, now codified, that distribution of a component of a patented device will not violate the patent if it is suitable for use in other ways. 35 U. S. C. § 271(c); Aro Mfg. Co. v. Convertible Top Replacement Co., 377 U. S. 476, 485 (1964) (noting codification of cases); id., at 486, n. 6 (same). The doctrine was devised to identify instances in which it may be presumed from distribution of an article in commerce that the distributor intended the article to be used to infringe another's patent, and so may justly be held liable for that infringement. "One who makes and sells articles which are only adapted to be used in a patented combination will be presumed to intend the natural consequences of his acts; he will be presumed to intend that they shall be used in the combination of the patent." New York Scaffolding Co. v. Whitney, 224 F. 452, 459 (CA8 1915); see also James Heekin Co. v. Baker, 138 F. 63, 66 (CA8 1905); Canda v. Michigan Malleable Iron Co., 124 F. 486, 489 (CA6 1903); Thomson-Houston Electric Co. v. Ohio Brass Co., 80 F. 712, 720-721 (CA6 1897); Red Jacket Mfg. Co. v. Davis, 82 F. 432, 439 (CA7 1897); Holly v. Vergennes Machine Co., 4 F. 74, 82 (CC Vt. 1880); Renwick v. Pond, 20 F. Cas. 536, 541 (No. 11,702) (CC SDNY 1872).

In sum, where an article is "good for nothing else" but infringement, Canda v. Michigan Malleable Iron Co., supra, at 489, there is no legitimate public interest in its unlicensed availability, and there is no injustice in presuming or imputing an intent to infringe, see Henry v. A. B. Dick Co., 224 U. S. 1, 48 (1912), overruled on other grounds, Motion Picture Patents Co. v. Universal Film Mfg. Co., 243 U. S. 502 (1917). Conversely, the doctrine absolves the equivocal conduct of selling an item with substantial lawful as well as unlawful uses, and limits liability to instances of more acute [545 U.S. 933] fault than the mere understanding that some of one's products will be misused. It leaves breathing room for innovation and a vigorous commerce. See Sony Corp. v. Universal City Studios, 464 U. S., at 442; Dawson Chemical Co. v. Rohm & Haas Co., 448 U. S. 176, 221 (1980); Henry v. A. B. Dick Co., supra, at 48.

The parties and many of the amici in this case think the key to resolving it is the Sony rule and, in particular, what it means for a product to be "capable of commercially significant noninfringing uses." Sony Corp. v. Universal City Studios, supra, at 442. MGM advances the argument that granting summary judgment to Grokster and StreamCast as to their current activities gave too much weight to the value of innovative technology, and too little to the copyrights infringed by users of their software, given that 90% of works available on one of the networks was shown to be copyrighted. Assuming the remaining 10% to be its noninfringing use, MGM says this should not qualify as "substantial," and the Court should quantify Sony to the extent of holding that a product used "principally" for infringement does not qualify. See Brief for Motion Picture Studio and Recording Company Petitioners 31. As mentioned before, Grokster and StreamCast reply by citing evidence that their software can be used to reproduce public domain works, and they point to copyright holders who actually encourage copying. Even if infringement is the principal practice with their software today, they argue, the noninfringing uses are significant and will grow.

We agree with MGM that the Court of Appeals misapplied Sony, which it read as limiting secondary liability quite beyond the circumstances to which the case applied. Sony barred secondary liability based on presuming or imputing intent to cause infringement solely from the design or distribution of a product capable of substantial lawful use, which the distributor knows is in fact used for infringement. The [545 U.S. 934] Ninth Circuit has read Sony's limitation to mean that whenever a product is capable of substantial lawful use, the producer can never be held contributorily liable for third parties' infringing use of it; it read the rule as being this broad, even when an actual purpose to cause infringing use is shown by evidence independent of design and distribution of the product, unless the distributors had "specific knowledge of infringement at a time at which they contributed to the infringement, and failed to act upon that information." 380 F. 3d, at 1162 (internal quotation marks and brackets omitted). Because the Circuit found the StreamCast and Grokster software capable of substantial lawful use, it concluded on the basis of its reading of Sony that neither company could be held liable, since there was no showing that their software, being without any central server, afforded them knowledge of specific unlawful uses.

This view of Sony, however, was error, converting the case from one about liability resting on imputed intent to one about liability on any theory. Because Sony did not displace other theories of secondary liability, and because we find below that it was error to grant summary judgment to the companies on MGM's inducement claim, we do not revisit Sony further, as MGM requests, to add a more quantified description of the point of balance between protection and commerce when liability rests solely on distribution with knowledge that unlawful use will occur. It is enough to note that the Ninth Circuit's judgment rested on an erroneous understanding of Sony and to leave further consideration of the Sony rule for a day when that may be required.

Sony's rule limits imputing culpable intent as a matter of law from the characteristics or uses of a distributed product. But nothing in Sony requires courts to ignore evidence of intent if there is such evidence, and the case was never meant to foreclose rules of fault-based liability derived from [545 U.S. 935] the common law.^[10] Sony Corp. v. Universal City Studios, supra, at 439 ("If vicarious liability is to be imposed on Sony in this case, it must rest on the fact that it has sold equipment with constructive knowledge" of the potential for infringement). Thus, where evidence goes beyond a product's characteristics or the knowledge that it may be put to infringing uses, and shows statements or actions directed to promoting infringement, Sony's staple-article rule will not preclude liability.

The classic case of direct evidence of unlawful purpose occurs when one induces commission of infringement by another, or "entic[es] or persuad[es] another" to infringe, Black's Law Dictionary 790 (8th ed. 2004), as by advertising. Thus at common law a copyright or patent defendant who "not only expected but invoked [infringing use] by advertisement" was liable for infringement "on principles recognized in every part of the law." Kalem Co. v. Harper Brothers, 222 U. S., at 62-63 (copyright infringement). See also Henry v. A. B. Dick Co., 224 U. S., at 48-49 (contributory liability for patent infringement may be found where a good's "most conspicuous use is one which will coöperate in an infringement when sale to such user is invoked by advertisement" of the infringing use); Thomson-Houston Electric Co. v. Kelsey Electric R. Specialty Co., 75 F. 1005, 1007-1008 (CA2 1896) (relying on advertisements and displays to find defendant's "willingness . . . to aid other persons in any attempts which they may be disposed to make towards [patent] infringement"); Rumford Chemical Works v. Hecker, 20 F. Cas. 1342, 1346 (No. 12,133) (CC NJ 1876) (demonstrations of infringing activity along with "avowals of the [infringing] purpose and use for which it was made" supported liability for patent infringement).

[545 U.S. 936] The rule on inducement of infringement as developed in the early cases is no different today.^[11] Evidence of "active steps . . . taken to encourage direct infringement," Oak Industries, Inc. v. Zenith Electronics Corp., 697 F. Supp. 988, 992 (ND Ill. 1988), such as advertising an infringing use or instructing how to engage in an infringing use, show an affirmative intent that the product be used to infringe, and a showing that infringement was encouraged overcomes the law's reluctance to find liability when a defendant merely sells a commercial product suitable for some lawful use, see, e. g., Water Technologies Corp. v. Calco, Ltd., 850 F. 2d 660, 668 (CA Fed. 1988) (liability for inducement where one "actively and knowingly aid[s] and abet[s] another's direct infringement" (emphasis deleted)); Fromberg, Inc. v. Thornhill, 315 F. 2d 407, 412-413 (CA5 1963) (demonstrations by sales staff of infringing uses supported liability for inducement); Haworth Inc. v. Herman Miller Inc., 37 USPQ 2d 1080, 1090 (WD Mich. 1994) (evidence that defendant "demonstrate[d] and recommend[ed] infringing configurations" of its product could support inducement liability); Sims v. Mack Trucks, Inc., 459 F. Supp. 1198, 1215 (ED Pa. 1978) (finding inducement where the use "depicted by the defendant in its promotional film and brochures infringes the . . . patent"), overruled on other grounds, 608 F. 2d 87 (CA3 1979). Cf. W. Keeton, D. Dobbs, R. Keeton, & D. Owen, Prosser and Keeton on Law of Torts 37 (5th ed. 1984) ("There is a definite tendency to impose greater responsibility upon a defendant whose conduct was intended to do harm, or was morally wrong").

For the same reasons that Sony took the staple-article doctrine of patent law as a model for its copyright safe-harbor rule, the inducement rule, too, is a sensible one for copyright. We adopt it here, holding that one who distributes a device with the object of promoting its use to infringe copyright, as [545 U.S. 937] shown by clear expression or other affirmative steps taken to foster infringement, is liable for the resulting acts of infringement by third parties. We are, of course, mindful of the need to keep from trenching on regular commerce or discouraging the development of technologies with lawful and unlawful potential. Accordingly, just as Sony did not find intentional inducement despite the knowledge of the VCR manufacturer that its device could be used to infringe, 464 U. S., at 439, n. 19, mere knowledge of infringing potential or of actual infringing uses would not be enough here to subject a distributor to liability. Nor would ordinary acts incident to product distribution, such as offering customers technical support or product updates, support liability in themselves. The inducement rule, instead, premises liability on purposeful, culpable expression and conduct, and thus does nothing to compromise legitimate commerce or discourage innovation having a lawful promise.

The only apparent question about treating MGM's evidence as sufficient to withstand summary judgment under the theory of inducement goes to the need on MGM's part to adduce evidence that StreamCast and Grokster communicated an inducing message to their software users. The classic instance of inducement is by advertisement or solicitation that broadcasts a message designed to stimulate others to commit violations. MGM claims that such a message is shown here. It is undisputed that StreamCast beamed onto the computer screens of users of Napster-compatible programs ads urging the adoption of its OpenNap program, which was designed, as its name implied, to invite the custom of patrons of Napster, then under attack in the courts for facilitating massive infringement. Those who accepted StreamCast's OpenNap program were offered software to perform the same services, which a factfinder could conclude [545 U.S. 938] would readily have been understood in the Napster market as the ability to download copyrighted music files. Grokster distributed an electronic newsletter containing links to articles promoting its software's ability to access popular copyrighted music. And anyone whose Napster or free file-sharing searches turned up a link to Grokster would have understood Grokster to be offering the same file-sharing ability as Napster, and to the same people who probably used Napster for infringing downloads; that would also have been the understanding of anyone offered Grokster's suggestively named Swaptor software, its version of OpenNap. And both companies communicated a clear message by responding affirmatively to requests for help in locating and playing copyrighted materials.

In StreamCast's case, of course, the evidence just described was supplemented by other unequivocal indications of unlawful purpose in the internal communications and advertising designs aimed at Napster users ("When the lights went off at Napster . . . where did the users go?" App. 836 (ellipsis in original)). Whether the messages were communicated is not to the point on this record. The function of the message in the theory of inducement is to prove by a defendant's own statements that his unlawful purpose disqualifies him from claiming protection (and incidentally to point to actual violators likely to be found among those who hear or read the message). See supra, at 935-937. Proving that a message was sent out, then, is the preeminent but not exclusive way of showing that active steps were taken with the purpose of bringing about infringing acts, and of showing that infringing acts took place by using the device distributed. Here, the summary judgment record is replete with other evidence that Grokster and StreamCast, unlike the manufacturer and distributor in Sony, acted with a purpose to cause copyright violations by use of software suitable for illegal use. See supra, at 924-927.

[545 U.S. 939] Three features of this evidence of intent are particularly notable. First, each company showed itself to be aiming to satisfy a known source of demand for copyright infringement, the market comprising former Napster users. StreamCast's internal documents made constant reference to Napster, it initially distributed its Morpheus software through an OpenNap program compatible with Napster, it advertised its OpenNap program to Napster users, and its Morpheus software functions as Napster did except that it could be used to distribute more kinds of files, including copyrighted movies and software programs. Grokster's name is apparently derived from Napster, it too initially offered an OpenNap program, its software's function is likewise comparable to Napster's, and it attempted to divert queries for Napster onto its own Web site. Grokster and StreamCast's efforts to supply services to former Napster users, deprived of a mechanism to copy and distribute what were overwhelmingly infringing files, indicate a principal, if not exclusive, intent on the part of each to bring about infringement.

Second, this evidence of unlawful objective is given added significance by MGM's showing that neither company attempted to develop filtering tools or other mechanisms to diminish the infringing activity using their software. While the Ninth Circuit treated the defendants' failure to develop such tools as irrelevant because they lacked an independent duty to monitor their users' activity, we think this evidence underscores Grokster's and StreamCast's intentional facilitation of their users' infringement.^[12]

Third, there is a further complement to the direct evidence of unlawful objective. It is useful to recall that StreamCast [545 U.S. 940] and Grokster make money by selling advertising space, by directing ads to the screens of computers employing their software. As the record shows, the more the software is used, the more ads are sent out and the greater the advertising revenue becomes. Since the extent of the software's use determines the gain to the distributors, the commercial sense of their enterprise turns on high-volume use, which the record shows is infringing.^[13] This evidence alone would not justify an inference of unlawful intent, but viewed in the context of the entire record its import is clear.

The unlawful objective is unmistakable.

In addition to intent to bring about infringement and distribution of a device suitable for infringing use, the inducement theory of course requires evidence of actual infringement by recipients of the device, the software in this case. As the account of the facts indicates, there is evidence of infringement on a gigantic scale, and there is no serious issue of the adequacy of MGM's showing on this point in order to survive the companies' summary judgment requests. Although [545 U.S. 941] an exact calculation of infringing use, as a basis for a claim of damages, is subject to dispute, there is no question that the summary judgment evidence is at least adequate to entitle MGM to go forward with claims for damages and equitable relief.

* * *

In sum, this case is significantly different from Sony and reliance on that case to rule in favor of StreamCast and Grokster was error. Sony dealt with a claim of liability based solely on distributing a product with alternative lawful and unlawful uses, with knowledge that some users would follow the unlawful course. The case struck a balance between the interests of protection and innovation by holding that the product's capability of substantial lawful employment should bar the imputation of fault and consequent secondary liability for the unlawful acts of others.

MGM's evidence in this case most obviously addresses a different basis of liability for distributing a product open to alternative uses. Here, evidence of the distributors' words and deeds going beyond distribution as such shows a purpose to cause and profit from third-party acts of copyright infringement. If liability for inducing infringement is ultimately found, it will not be on the basis of presuming or imputing fault, but from inferring a patently illegal objective from statements and actions showing what that objective was.

There is substantial evidence in MGM's favor on all elements of inducement, and summary judgment in favor of Grokster and StreamCast was error. On remand, reconsideration of MGM's motion for summary judgment will be in order.

The judgment of the Court of Appeals is vacated, and the case is remanded for further proceedings consistent with this opinion.

It is so ordered.

[545 U.S. 942] JUSTICE GINSBURG, with whom THE CHIEF JUSTICE and JUSTICE KENNEDY join, concurring.

I concur in the Court's decision, which vacates in full the judgment of the Court of Appeals for the Ninth Circuit, ante, at 941, and write separately to clarify why I conclude that the Court of Appeals misperceived, and hence misapplied, our holding in Sony Corp. of America v. Universal City Studios, Inc., 464 U. S. 417 (1984). There is here at least a "genuine issue as to [a] material fact," Fed. Rule Civ. Proc. 56(c), on the liability of Grokster or StreamCast, not only for actively inducing copyright infringement, but also, or alternatively, based on the distribution of their software products, for contributory copyright infringement. On neither score was summary judgment for Grokster and StreamCast warranted.

At bottom, however labeled, the question in this case is whether Grokster and StreamCast are liable for the direct infringing acts of others. Liability under our jurisprudence may be predicated on actively encouraging (or inducing) infringement through specific acts (as the Court's opinion develops) or on distributing a product distributees use to infringe copyrights, if the product is not capable of "substantial" or "commercially significant" noninfringing uses. Sony, 464 U. S., at 442; see also 3 M. Nimmer & D. Nimmer, Nimmer on Copyright § 12.04[A][2] (2005). While the two categories overlap, they capture different culpable behavior. Long coexisting, both are now codified in patent law. Compare 35 U. S. C. § 271(b) (active inducement liability) with § 271(c) (contributory liability for distribution of a product not "suitable for substantial noninfringing use").

In Sony, 464 U. S. 417, the Court considered Sony's liability for selling the Betamax videocassette recorder. It did so enlightened by a full trial record. Drawing an analogy to the staple article of commerce doctrine from patent law, [545 U.S. 943] the Sony Court observed that the "sale of an article . . . adapted to [a patent] infringing use" does not suffice "to make the seller a contributory infringer" if the article "is also adapted to other and lawful uses." Id., at 441 (quoting Henry v. A. B. Dick Co., 224 U. S. 1, 48 (1912), overruled on other grounds, Motion Picture Patents Co. v. Universal Film Mfg. Co., 243 U. S. 502, 517 (1917)).

"The staple article of commerce doctrine" applied to copyright, the Court stated, "must strike a balance between a copyright holder's legitimate demand for effective—not merely symbolic—protection of the statutory monopoly, and the rights of others freely to engage in substantially unrelated areas of commerce." Sony, 464 U. S., at 442. "Accordingly," the Court held, "the sale of copying equipment, like the sale of other articles of commerce, does not constitute contributory infringement if the product is widely used for legitimate, unobjectionable purposes. Indeed, it need merely be capable of substantial noninfringing uses." Ibid. Thus, to resolve the Sony case, the Court explained, it had to determine "whether the Betamax is capable of commercially significant noninfringing uses." Ibid.

To answer that question, the Court considered whether "a significant number of [potential uses of the Betamax were] noninfringing." Ibid. The Court homed in on one potential use—private, noncommercial time-shifting of television programs in the home (i. e., recording a broadcast TV program for later personal viewing). Time-shifting was noninfringing, the Court concluded, because in some cases trial testimony showed it was authorized by the copyright holder, id., at 443-447, and in others it qualified as legitimate fair use, id., at 447-455. Most purchasers used the Betamax principally to engage in time-shifting, id., at 421, 423, a use that "plainly satisfie[d]" the Court's standard, id., at 442. Thus, there was no need in Sony to "give precise content to the question of how much [actual or potential] use is commercially [545 U.S. 944] significant." Ibid.^[14] Further development was left for later days and cases.

The Ninth Circuit went astray, I will endeavor to explain, when that court granted summary judgment to Grokster and StreamCast on the charge of contributory liability based on distribution of their software products. Relying on its earlier opinion in A&M Records, Inc. v. Napster, Inc., 239 F. 3d 1004 (CA9 2001), the Court of Appeals held that "if substantial noninfringing use was shown, the copyright owner would be required to show that the defendant had reasonable knowledge of specific infringing files." 380 F. 3d 1154, 1161 (CA9 2004). "A careful examination of the record," the [545 U.S. 945] court concluded, "indicates that there is no genuine issue of material fact as to noninfringing use." Ibid. The appeals court pointed to the band Wilco, which made one of its albums available for free downloading, to other recording artists who may have authorized free distribution of their music through the Internet, and to public domain literary works and films available through Grokster's and StreamCast's software. Ibid. Although it acknowledged petitioners' (hereinafter MGM) assertion that "the vast majority of the software use is for copyright infringement," the court concluded that Grokster's and StreamCast's proffered evidence met Sony's requirement that "a product need only be capable of substantial noninfringing uses." 380 F. 3d, at 1162.^[15]

This case differs markedly from Sony. Cf. Peters, Brace Memorial Lecture: Copyright Enters the Public Domain, 51 J. Copyright Soc. 701, 724 (2004) ("The Grokster panel's reading of Sony is the broadest that any court has given it . . . ."). Here, there has been no finding of any fair use and little beyond anecdotal evidence of noninfringing uses. In finding the Grokster and StreamCast software products capable of substantial noninfringing uses, the District Court and the Court of Appeals appear to have relied largely on declarations submitted by the defendants. These declarations include assertions (some of them hearsay) that a number of copyright owners authorize distribution of their works on the Internet and that some public domain material is available through peer-to-peer networks including those accessed through Grokster's and StreamCast's software. 380 F. 3d, at 1161; 259 F. Supp. 2d 1029, 1035-1036 (CD Cal. 2003); App. 125-171.

[545 U.S. 946] The District Court declared it "undisputed that there are substantial noninfringing uses for Defendants' software," thus obviating the need for further proceedings. 259 F. Supp. 2d, at 1035. This conclusion appears to rest almost entirely on the collection of declarations submitted by Grokster and StreamCast. Ibid. Review of these declarations reveals mostly anecdotal evidence, sometimes obtained secondhand, of authorized copyrighted works or public domain works available online and shared through peer-to-peer networks, and general statements about the benefits of peer-to-peer technology. See, e. g., Decl. of Janis Ian ¶ 13, App. 128 ("P2P technologies offer musicians an alternative channel for promotion and distribution."); Decl. of Gregory Newby ¶ 12, id., at 136 ("Numerous authorized and public domain Project Gutenberg eBooks are made available on Morpheus, Kazaa, Gnutella, Grokster, and similar software products."); Decl. of Aram Sinnreich ¶ 6, id., at 151 ("file sharing seems to have a net positive impact on music sales"); Decl. of John Busher ¶ 8, id., at 166 ("I estimate that Acoustica generates sales of between $1,000 and $10,000 per month as a result of the distribution of its trialware software through the Gnutella and FastTrack Networks."); Decl. of Patricia D. Hoekman ¶¶ 3-4, id., at 169-170 (search on Morpheus for "President Bush speeches" found several video recordings, searches for "Declaration of Independence" and "Bible" found various documents and declarant was able to download a copy of the Declaration); Decl. of Sean L. Mayers ¶ 11, id., at 67 ("Existing open, decentralized peer-to-peer file-sharing networks . . . offer content owners distinct business advantages over alternate online distribution technologies."). Compare Decl. of Brewster Kahle ¶ 20, id., at 142 ("Those who download the Prelinger films . . . are entitled to redistribute those files, and the Archive welcomes their redistribution by the Morpheus-Grokster-KaZaa community of users."), with Deposition of Brewster Kahle (Sept. 18, [545 U.S. 947] 2002), id., at 396-403 (testifying that he has no knowledge of any person downloading a Prelinger film using Morpheus, Grokster, or KaZaA). Compare also Decl. of Richard Prelinger ¶ 17, id., at 147 ("[W]e welcome further redistribution of the Prelinger films . . . by individuals using peer-to-peer software products like Morpheus, KaZaA and Grokster."), with Deposition of Richard Prelinger (Oct. 1, 2002), id., at 410-411 ("Q. What is your understanding of Grokster? A. I have no understanding of Grokster. . . . Q. Do you know whether any user of the Grokster software has made available to share any Prelinger film? A. No."). See also Deposition of Aram Sinnreich (Sept. 25, 2002), id., at 390 (testimony about the band Wilco based on "[t]he press and industry news groups and scuttlebutt."). These declarations do not support summary judgment in the face of evidence, proffered by MGM, of overwhelming use of Grokster's and StreamCast's software for infringement.^[16]

[545 U.S. 948] Even if the absolute number of noninfringing files copied using the Grokster and StreamCast software is large, it does not follow that the products are therefore put to substantial noninfringing uses and are thus immune from liability. The number of noninfringing copies may be reflective of, and dwarfed by, the huge total volume of files shared. Further, the District Court and the Court of Appeals did not sharply distinguish between uses of Grokster's and StreamCast's software products (which this case is about) and uses of peer-to-peer technology generally (which this case is not about).

In sum, when the record in this case was developed, there was evidence that Grokster's and StreamCast's products were, and had been for some time, overwhelmingly used to infringe, ante, at 922-924; App. 434-439, 476-481, and that this infringement was the overwhelming source of revenue from the products, ante, at 925-926; 259 F. Supp. 2d, at 1043-1044. Fairly appraised, the evidence was insufficient to demonstrate, beyond genuine debate, a reasonable prospect that substantial or commercially significant noninfringing uses were likely to develop over time. On this record, the District Court should not have ruled dispositively on the contributory infringement charge by granting summary judgment to Grokster and StreamCast.^[17]

If, on remand, the case is not resolved on summary judgment in favor of MGM based on Grokster and StreamCast actively inducing infringement, the Court of Appeals, I [545 U.S. 949] would emphasize, should reconsider, on a fuller record, its interpretation of Sony's product distribution holding.

---------------

JUSTICE BREYER, with whom JUSTICE STEVENS and JUSTICE O'CONNOR join, concurring.

I agree with the Court that the distributor of a dual-use technology may be liable for the infringing activities of third parties where he or she actively seeks to advance the infringement. Ante, at 919. I further agree that, in light of our holding today, we need not now "revisit" Sony Corp. of America v. Universal City Studios, Inc., 464 U. S. 417 (1984). Ante, at 934. Other Members of the Court, however, take up the Sony question: whether Grokster's product is "capable of `substantial' or `commercially significant' noninfringing uses." Ante, at 942 (GINSBURG, J., concurring) (quoting Sony, supra, at 442). And they answer that question by stating that the Court of Appeals was wrong when it granted summary judgment on the issue in Grokster's favor. Ante, at 944. I write to explain why I disagree with them on this matter.

The Court's opinion in Sony and the record evidence (as described and analyzed in the many briefs before us) together convince me that the Court of Appeals' conclusion has adequate legal support.

I begin with Sony's standard. In Sony, the Court considered the potential copyright liability of a company that did not itself illegally copy protected material, but rather sold a machine—a videocassette recorder (VCR)—that could be used to do so. A buyer could use that machine for non-infringing purposes, such as recording for later viewing (sometimes called "`time-shifting,'" Sony, 464 U. S., at 421) uncopyrighted television programs or copyrighted programs with a copyright holder's permission. The buyer could use [545 U.S. 950] the machine for infringing purposes as well, such as building libraries of taped copyrighted programs. Or, the buyer might use the machine to record copyrighted programs under circumstances in which the legal status of the act of recording was uncertain (i. e., where the copying may, or may not, have constituted a "fair use," id., at 425-426). Sony knew many customers would use its VCRs to engage in unauthorized copying and "`library-building.'" Id., at 458-459 (Blackmun, J., dissenting). But that fact, said the Court, was insufficient to make Sony itself an infringer. And the Court ultimately held that Sony was not liable for its customers' acts of infringement.

In reaching this conclusion, the Court recognized the need for the law, in fixing secondary copyright liability, to "strike a balance between a copyright holder's legitimate demand for effective—not merely symbolic—protection of the statutory monopoly, and the rights of others freely to engage in substantially unrelated areas of commerce." Id., at 442. It pointed to patent law's "staple article of commerce" doctrine, ibid., under which a distributor of a product is not liable for patent infringement by its customers unless that product is "unsuited for any commercial noninfringing use." Dawson Chemical Co. v. Rohm & Haas Co., 448 U. S. 176, 198 (1980). The Court wrote that the sale of copying equipment, "like the sale of other articles of commerce, does not constitute contributory infringement if the product is widely used for legitimate, unobjectionable purposes. Indeed, it need merely be capable of substantial noninfringing uses." Sony, 464 U. S., at 442 (emphasis added). The Court ultimately characterized the legal "question" in the particular case as "whether [Sony's VCR] is capable of commercially significant noninfringing uses" (while declining to give "precise content" to these terms). Ibid. (emphasis added).

It then applied this standard. The Court had before it a survey (commissioned by the District Court and then prepared by the respondents) showing that roughly 9% of all [545 U.S. 951] VCR recordings were of the type—namely, religious, educational, and sports programming—owned by producers and distributors testifying on Sony's behalf who did not object to time-shifting. See Brief for Respondents, O. T. 1983, No. 81-1687, pp. 52-53; see also Sony, supra, at 424 (7.3% of all Sony VCR use is to record sports programs; representatives of the sports leagues do not object). A much higher percentage of VCR users had at one point taped an authorized program, in addition to taping unauthorized programs. And the plaintiffs—not a large class of content providers as in this case—owned only a small percentage of the total available unauthorized programming. See ante, at 947, n. 3 (GINSBURG, J., concurring). But of all the taping actually done by Sony's customers, only around 9% was of the sort the Court referred to as authorized.

The Court found that the magnitude of authorized programming was "significant," and it also noted the "significant potential for future authorized copying." 464 U. S., at 444. The Court supported this conclusion by referencing the trial testimony of professional sports league officials and a religious broadcasting representative. Id., at 444, and n. 24. It also discussed (1) a Los Angeles educational station affiliated with the Public Broadcasting Service that made many of its programs available for home taping, and (2) Mr. Rogers' Neighborhood, a widely watched children's program. Id., at 445. On the basis of this testimony and other similar evidence, the Court determined that producers of this kind had authorized duplication of their copyrighted programs "in significant enough numbers to create a substantial market for a noninfringing use of the" VCR. Id., at 447, n. 28 (emphasis added).

The Court, in using the key word "substantial," indicated that these circumstances alone constituted a sufficient basis for rejecting the imposition of secondary liability. See id., at 456 ("Sony demonstrated a significant likelihood that substantial numbers of copyright holders" would not object [545 U.S. 952] to time-shifting (emphasis added)). Nonetheless, the Court buttressed its conclusion by finding separately that, in any event, un-authorized time-shifting often constituted not infringement, but "fair use." Id., at 447-456.

When measured against Sony's underlying evidence and analysis, the evidence now before us shows that Grokster passes Sony's test—that is, whether the company's product is capable of substantial or commercially significant noninfringing uses. Id., at 442. For one thing, petitioners' (hereinafter MGM) own expert declared that 75% of current files available on Grokster are infringing and 15% are "likely infringing." See App. 436-439, ¶¶ 6-17 (Decl. of Dr. Ingram Olkin); cf. ante, at 922 (opinion of the Court). That leaves some number of files near 10% that apparently are noninfringing, a figure very similar to the 9% or so of authorized time-shifting uses of the VCR that the Court faced in Sony.

As in Sony, witnesses here explained the nature of the noninfringing files on Grokster's network without detailed quantification. Those files include:

—Authorized copies of music by artists such as Wilco, Janis Ian, Pearl Jam, Dave Matthews, John Mayer, and others. See App. 152-153, ¶¶ 9-13 (Decl. of Aram Sinnreich) (Wilco's "lesson has already been adopted by artists still signed to their major labels"); id., at 170, ¶¶ 5-7 (Decl. of Patricia D. Hoekman) (locating "numerous audio recordings" that were authorized for swapping); id., at 74, ¶ 10 (Decl. of Daniel B. Rung) (describing Grokster's partnership with a company that hosts music from thousands of independent artists)

—Free electronic books and other works from various online publishers, including Project Gutenberg. See id., at 136, ¶ 12 (Decl. of Gregory Newby) ("Numerous authorized and public domain Project Gutenberg eBooks are made available" on Grokster. Project Gutenberg "welcomes this widespread [545 U.S. 953] sharing . . . using these software products[,] since they assist us in meeting our objectives"); id., at 159-160, ¶ 32 (Decl. of Sinnreich)

—Public domain and authorized software, such as WinZip 8.1. Id., at 170, ¶ 8 (Decl. of Hoekman); id., at 165, ¶¶ 4-7 (Decl. of John Busher)

—Licensed music videos and television and movie segments distributed via digital video packaging with the permission of the copyright holder. Id., at 70, ¶ 24 (Decl. of Sean L. Mayers).

The nature of these and other lawfully swapped files is such that it is reasonable to infer quantities of current lawful use roughly approximate to those at issue in Sony. At least, MGM has offered no evidence sufficient to survive summary judgment that could plausibly demonstrate a significant quantitative difference. See ante, at 922 (opinion of the Court); see also Brief for Motion Picture Studio and Recording Company Petitioners i (referring to "at least 90% of the total use of the services"); but see ante, at 947, n. 3 (GINSBURG, J., concurring). To be sure, in quantitative terms these uses account for only a small percentage of the total number of uses of Grokster's product. But the same was true in Sony, which characterized the relatively limited authorized copying market as "substantial." (The Court made clear as well in Sony that the amount of material then presently available for lawful copying—if not actually copied— was significant, see 464 U. S., at 444, and the same is certainly true in this case.)

Importantly, Sony also used the word "capable," asking whether the product is "capable of" substantial noninfringing uses. Its language and analysis suggest that a figure like 10%, if fixed for all time, might well prove insufficient, but that such a figure serves as an adequate foundation where there is a reasonable prospect of expanded legitimate uses over time. See ibid. (noting a "significant potential for future authorized copying"). And its language also indicates [545 U.S. 954] the appropriateness of looking to potential future uses of the product to determine its "capability."

Here the record reveals a significant future market for noninfringing uses of Grokster-type peer-to-peer software. Such software permits the exchange of any sort of digital file—whether that file does, or does not, contain copyrighted material. As more and more uncopyrighted information is stored in swappable form, it seems a likely inference that lawful peer-to-peer sharing will become increasingly prevalent. See, e. g., App. 142, ¶ 20 (Decl. of Brewster Kahle) ("[T]he [Internet Archive] welcomes [the] redistribution [of authorized films] by the Morpheus-Grokster-KaZaa community of users"); id., at 166, ¶ 8 (Decl. of Busher) (sales figures of $1,000 to $10,000 per month through peer-to-peer networks "will increase in the future as Acoustica's trialware is more widely distributed through these networks"); id., at 156-163, ¶¶ 21-40 (Decl. of Sinnreich).

And that is just what is happening. Such legitimate noninfringing uses are coming to include the swapping of: research information (the initial purpose of many peer-to-peer networks); public domain films (e. g., those owned by the Prelinger Archive); historical recordings and digital educational materials (e. g., those stored on the Internet Archive); digital photos (OurPictures, for example, is starting a P2P photo-swapping service); "shareware" and "freeware" (e. g., Linux and certain Windows software); secure licensed music and movie files (Intent MediaWorks, for example, protects licensed content sent across P2P networks); news broadcasts past and present (the BBC Creative Archive lets users "rip, mix and share the BBC"); user-created audio and video files (including "podcasts" that may be distributed through P2P software); and all manner of free "open content" works collected by Creative Commons (one can search for Creative Commons material on StreamCast). See Brief for Distributed Computing Industry Association as Amicus Curiae 15-26; Merges, A New Dynamism in the Public Domain, 71 [545 U.S. 955] U. Chi. L. Rev. 183 (2004). I can find nothing in the record that suggests that this course of events will not continue to flow naturally as a consequence of the character of the software taken together with the foreseeable development of the Internet and of information technology. Cf. ante, at 920 (opinion of the Court) (discussing the significant benefits of peer-to-peer technology).

There may be other now-unforeseen noninfringing uses that develop for peer-to-peer software, just as the homevideo rental industry (unmentioned in Sony) developed for the VCR. But the foreseeable development of such uses, when taken together with an estimated 10% noninfringing material, is sufficient to meet Sony's standard. And while Sony considered the record following a trial, there are no facts asserted by MGM in its summary judgment filings that lead me to believe the outcome after a trial here could be any different. The lower courts reached the same conclusion.

Of course, Grokster itself may not want to develop these other noninfringing uses. But Sony's standard seeks to protect not the Groksters of this world (which in any event may well be liable under today's holding), but the development of technology more generally. And Grokster's desires in this respect are beside the point.

The real question here, I believe, is not whether the record evidence satisfies Sony. As I have interpreted the standard set forth in that case, it does. And of the Courts of Appeals that have considered the matter, only one has proposed interpreting Sony more strictly than I would do—in a case where the product might have failed under any standard. In re Aimster Copyright Litigation, 334 F. 3d 643, 653 (CA7 2003) (defendant "failed to show that its service is ever used for any purpose other than to infringe" copyrights (emphasis added)); see Matthew Bender & Co. v. West Pub. Co., 158 [545 U.S. 956] F. 3d 693, 706-707 (CA2 1998) (court did not require that noninfringing uses be "predominant," it merely found that they were predominant, and therefore provided no analysis of Sony's boundaries); but see ante, at 944, n. 1 (GINSBURG, J., concurring); see also A&M Records, Inc. v. Napster, Inc., 239 F. 3d 1004, 1020 (CA9 2001) (discussing Sony); Cable/Home Communication Corp. v. Network Productions, Inc., 902 F. 2d 829, 842-847 (CA11 1990) (same); Vault Corp. v. Quaid Software, Ltd., 847 F. 2d 255, 262 (CA5 1988) (same); cf. Dynacore Holdings Corp. v. U. S. Philips Corp., 363 F. 3d 1263, 1275 (CA Fed. 2004) (same); see also Doe v. GTE Corp., 347 F. 3d 655, 661 (CA7 2003) ("A person may be liable as a contributory infringer if the product or service it sells has no (or only slight) legal use").

Instead, the real question is whether we should modify the Sony standard, as MGM requests, or interpret Sony more strictly, as I believe JUSTICE GINSBURG'S approach would do in practice. Compare ante, at 944-948 (concurring opinion) (insufficient evidence in this case of both present lawful uses and of a reasonable prospect that substantial noninfringing uses would develop over time), with Sony, 464 U. S., at 442-447 (basing conclusion as to the likely existence of a substantial market for authorized copying upon general declarations, some survey data, and common sense).

As I have said, Sony itself sought to "strike a balance between a copyright holder's legitimate demand for effective— not merely symbolic—protection of the statutory monopoly, and the rights of others freely to engage in substantially unrelated areas of commerce." Id., at 442. Thus, to determine whether modification, or a strict interpretation, of Sony is needed, I would ask whether MGM has shown that Sony incorrectly balanced copyright and new-technology interests. In particular: (1) Has Sony (as I interpret it) worked to protect new technology? (2) If so, would modification or strict interpretation significantly weaken that protection? (3) If [545 U.S. 957] so, would new or necessary copyright-related benefits outweigh any such weakening?

The first question is the easiest to answer. Sony's rule, as I interpret it, has provided entrepreneurs with needed assurance that they will be shielded from copyright liability as they bring valuable new technologies to market.

Sony's rule is clear. That clarity allows those who develop new products that are capable of substantial noninfringing uses to know, ex ante, that distribution of their product will not yield massive monetary liability. At the same time, it helps deter them from distributing products that have no other real function than—or that are specifically intended for—copyright infringement, deterrence that the Court's holding today reinforces (by adding a weapon to the copyright holder's legal arsenal).

Sony's rule is strongly technology protecting. The rule deliberately makes it difficult for courts to find secondary liability where new technology is at issue. It establishes that the law will not impose copyright liability upon the distributors of dual-use technologies (who do not themselves engage in unauthorized copying) unless the product in question will be used almost exclusively to infringe copyrights (or unless they actively induce infringements as we today describe). Sony thereby recognizes that the copyright laws are not intended to discourage or to control the emergence of new technologies, including (perhaps especially) those that help disseminate information and ideas more broadly or more efficiently. Thus Sony's rule shelters VCRs, typewriters, tape recorders, photocopiers, computers, cassette players, compact disc burners, digital video recorders, MP3 players, Internet search engines, and peer-to-peer software. But Sony's rule does not shelter descramblers, even if one could theoretically use a descrambler in a noninfringing way. 464 [545 U.S. 958] U. S., at 441-442. Compare Cable/Home Communication Corp., supra, at 837-850 (developer liable for advertising television signal descrambler), with Vault Corp., supra, at 262 (primary use infringing but a substantial noninfringing use).

Sony's rule is forward looking. It does not confine its scope to a static snapshot of a product's current uses (thereby threatening technologies that have undeveloped future markets). Rather, as the VCR example makes clear, a product's market can evolve dramatically over time. And Sony—by referring to a capacity for substantial noninfringing uses—recognizes that fact. Sony's word "capable" refers to a plausible, not simply a theoretical, likelihood that such uses will come to pass, and that fact anchors Sony in practical reality. Cf. Aimster, 334 F. 3d, at 651.

Sony's rule is mindful of the limitations facing judges where matters of technology are concerned. Judges have no specialized technical ability to answer questions about present or future technological feasibilility or commercial viability where technology professionals, engineers, and venture capitalists themselves may radically disagree and where answers may differ depending upon whether one focuses upon the time of product development or the time of distribution. Consider, for example, the question whether devices can be added to Grokster's software that will filter out infringing files. MGM tells us this is easy enough to do, as do several amici that produce and sell the filtering technology. See, e. g., Brief for Motion Picture Studio and Recording Company Petitioners 11; Brief for Audible Magic Corp. et al. as Amici Curiae 3-10. Grokster says it is not at all easy to do, and not an efficient solution in any event, and several apparently disinterested computer science professors agree. See Brief for Respondents 31; Brief for Computer Science Professor Harold Abelson et al. as Amici Curiae 6-10, 14-18. Which account should a judge credit? Sony says that the judge will not necessarily have to decide.

[545 U.S. 959] Given the nature of the Sony rule, it is not surprising that in the last 20 years, there have been relatively few contributory infringement suits—based on a product distribution theory—brought against technology providers (a small handful of federal appellate court cases and perhaps fewer than two dozen District Court cases in the last 20 years). I have found nothing in the briefs or the record that shows that Sony has failed to achieve its innovation-protecting objective.

The second, more difficult, question is whether a modified Sony rule (or a strict interpretation) would significantly weaken the law's ability to protect new technology. JUSTICE GINSBURG'S approach would require defendants to produce considerably more concrete evidence—more than was presented here—to earn Sony's shelter. That heavier evidentiary demand, and especially the more dramatic (case-by-case balancing) modifications that MGM and the Government seek, would, I believe, undercut the protection that Sony now offers.

To require defendants to provide, for example, detailed evidence—say, business plans, profitability estimates, projected technological modifications, and so forth—would doubtless make life easier for copyright holder plaintiffs. But it would simultaneously increase the legal uncertainty that surrounds the creation or development of a new technology capable of being put to infringing uses. Inventors and entrepreneurs (in the garage, the dorm room, the corporate lab, or the boardroom) would have to fear (and in many cases endure) costly and extensive trials when they create, produce, or distribute the sort of information technology that can be used for copyright infringement. They would often be left guessing as to how a court, upon later review of the product and its uses, would decide when necessarily rough estimates amounted to sufficient evidence. They would have no way to predict how courts would weigh the respective [545 U.S. 960] values of infringing and noninfringing uses; determine the efficiency and advisability of technological changes; or assess a product's potential future markets. The price of a wrong guess—even if it involves a good-faith effort to assess technical and commercial viability—could be large statutory damages (not less than $750 and up to $30,000 per infringed work). 17 U. S. C. § 504(c)(1). The additional risk and uncertainty would mean a consequent additional chill of technological development.

The third question—whether a positive copyright impact would outweigh any technology-related loss—I find the most difficult of the three. I do not doubt that a more intrusive Sony test would generally provide greater revenue security for copyright holders. But it is harder to conclude that the gains on the copyright swings would exceed the losses on the technology roundabouts.

For one thing, the law disfavors equating the two different kinds of gain and loss; rather, it leans in favor of protecting technology. As Sony itself makes clear, the producer of a technology which permits unlawful copying does not himself engage in unlawful copying—a fact that makes the attachment of copyright liability to the creation, production, or distribution of the technology an exceptional thing. See 464 U. S., at 431 (courts "must be circumspect" in construing the copyright laws to preclude distribution of new technologies). Moreover, Sony has been the law for some time. And that fact imposes a serious burden upon copyright holders like MGM to show a need for change in the current rules of the game, including a more strict interpretation of the test. See, e. g., Brief for Motion Picture Studio and Recording Company Petitioners 31 (Sony should not protect products when the "primary or principal" use is infringing).

In any event, the evidence now available does not, in my view, make out a sufficiently strong case for change. To say [545 U.S. 961] this is not to doubt the basic need to protect copyrighted material from infringement. The Constitution itself stresses the vital role that copyright plays in advancing the "useful Arts." Art. I, § 8, cl. 8. No one disputes that "reward to the author or artist serves to induce release to the public of the products of his creative genius." United States v. Paramount Pictures, Inc., 334 U. S. 131, 158 (1948). And deliberate unlawful copying is no less an unlawful taking of property than garden-variety theft. See, e. g., 18 U. S. C. § 2319 (2000 ed. and Supp. II) (criminal copyright infringement); § 1961(1)(B) (2000 ed., Supp. II) (copyright infringement can be a predicate act under the Racketeer Influenced and Corrupt Organizations Act); § 1956(c)(7)(D) (2000 ed., Supp. II) (money laundering includes the receipt of proceeds from copyright infringement). But these highly general principles cannot by themselves tell us how to balance the interests at issue in Sony or whether Sony's standard needs modification. And at certain key points, information is lacking.

Will an unmodified Sony lead to a significant diminution in the amount or quality of creative work produced? Since copyright's basic objective is creation and its revenue objectives but a means to that end, this is the underlying copyright question. See Twentieth Century Music Corp. v. Aiken, 422 U. S. 151, 156 (1975) ("Creative work is to be encouraged and rewarded, but private motivation must ultimately serve the cause of promoting broad public availability of literature, music, and the other arts"). And its answer is far from clear.

Unauthorized copying likely diminishes industry revenue, though it is not clear by how much. Compare S. Liebowitz, Will MP3 Downloads Annihilate the Record Industry? The Evidence So Far 2 (June 2003), http://www.utdallas.edu/~liebowit/intprop/records.pdf (all Internet materials as visited June 24, 2005, and available in Clerk of Court's case file)

[545 U.S. 962] (file sharing has caused a decline in music sales), and Press Release, Informa Telecoms & Media, Steady Download Growth Defies P2P (Dec. 6, 2004), http://www.informatm.com (citing Informa Media Group Report, Music on the Internet (5th ed. 2004)) (estimating total lost sales to the music industry in the range of $2 billion annually), with F. Oberholzer & K. Strumpf, The Effect of File Sharing on Record Sales: An Empirical Analysis 24 (Mar. 2004), www.unc.edu/~cigar/papers/FileSharing_March2004.pdf (academic study concluding that "file sharing has no statistically significant effect on purchases of the average album"), and D. McGuire, Study: File-Sharing No Threat to Music Sales (Mar. 29, 2004), http://www.washingtonpost.com/ac2/wp-dyn/A34300-2004Mar29?language=printer (discussing mixed evidence).

The extent to which related production has actually and resultingly declined remains uncertain, though there is good reason to believe that the decline, if any, is not substantial. See, e. g., M. Madden, Pew Internet & American Life Project, Artists, Musicians, and the Internet 21 (Dec. 5, 2004), http://www.pewinternet.org/pdfs/PIP_Artists. Musicians_Report.pdf (nearly 70% of musicians believe that file sharing is a minor threat or no threat at all to creative industries); Benkler, Sharing Nicely: On Shareable Goods and the Emergence of Sharing as a Modality of Economic Production, 114 Yale L. J. 273, 351-352 (2004) ("Much of the actual flow of revenue to artists—from performances and other sources—is stable even assuming a complete displacement of the CD market by peer-to-peer distribution . . . . [I]t would be silly to think that music, a cultural form without which no human society has existed, will cease to be in our world [because of illegal file swapping]").

More importantly, copyright holders at least potentially have other tools available to reduce piracy and to abate whatever threat it poses to creative production. As today's opinion makes clear, a copyright holder may proceed against [545 U.S. 963] a technology provider where a provable specific intent to infringe (of the kind the Court describes) is present. Ante, at 941. Services like Grokster may well be liable under an inducement theory.

In addition, a copyright holder has always had the legal authority to bring a traditional infringement suit against one who wrongfully copies. Indeed, since September 2003, the Recording Industry Association of America (RIAA) has filed "thousands of suits against people for sharing copyrighted material." Walker, New Movement Hits Universities: Get Legal Music, Washington Post, Mar. 17, 2005, p. E1. These suits have provided copyright holders with damages; have served as a teaching tool, making clear that much file sharing, if done without permission, is unlawful; and apparently have had a real and significant deterrent effect. See, e. g., L. Rainie, M. Madden, D. Hess, & G. Mudd, Pew Internet Project and comScore Media Metrix Data Memo: The state of music downloading and file-sharing online 2, 4, 6, 10 (Apr. 2004), http://www.pewinternet.org/pdfs/PIP_Filesharing_April_04.pdf (number of people downloading files fell from a peak of roughly 35 million to roughly 23 million in the year following the first suits; 38% of current downloaders report downloading fewer files because of the suits); M. Madden & L. Rainie, Pew Internet Project Data Memo: Music and video downloading moves beyond P2P, p. 7 (Mar. 2005), http://www. pewinternet.org/pdfs/PIP_Filesharing_March05.pdf (number of downloaders has "inched up" but "continues to rest well below the peak level"); Note, Costs and Benefits of the Recording Industry's Litigation Against Individuals, 20 Berkeley Tech. L. J. 571 (2005); but see Evangelista, File Sharing; Downloading Music and Movie Files is as Popular as Ever, San Francisco Chronicle, Mar. 28, 2005, p. E1 (referring to the continuing "tide of rampant copyright infringement," while noting that the RIAA says it believes the "campaign of lawsuits and public education has at least contained the problem").

[545 U.S. 964] Further, copyright holders may develop new technological devices that will help curb unlawful infringement. Some new technology, called "digital `watermarking'" and "digital fingerprint[ing]," can encode within the file information about the author and the copyright scope and date, which "fingerprints" can help to expose infringers. RIAA Reveals Method to Madness, Wired News (Aug. 28, 2003), http:// www.wired.com/news/digiwood/0,1412,60222,00.html; Besek, Anti-Circumvention Laws and Copyright: A Report from the Kernochan Center for Law, Media and the Arts, 27 Colum. J. L. & Arts 385, 391, 451 (2004). Other technology can, through encryption, potentially restrict users' ability to make a digital copy. See J. Borland, Tripping the Rippers, C/net News.com (Sept. 28, 2001), http://news.com.com/ Tripping+the+rippers/2009-1023_3-273619.html; but see Brief for Bridgemar Services, Ltd. d/b/a iMesh.com as Amicus Curiae 5-8 (arguing that peer-to-peer service providers can more easily block unlawful swapping).

At the same time, advances in technology have discouraged unlawful copying by making lawful copying (e. g., downloading music with the copyright holder's permission) cheaper and easier to achieve. Several services now sell music for less than $1 per song. (Walmart.com, for example, charges $0.88 each.) Consequently, many consumers initially attracted to the convenience and flexibility of services like Grokster are now migrating to lawful paid services (services with copying permission) where they can enjoy at little cost even greater convenience and flexibility without engaging in unlawful swapping. See Wu, When Code Isn't Law, 89 Va. L. Rev. 679, 731-735 (2003) (noting the prevalence of technological problems on unpaid swapping sites); K. Dean, P2P Tilts Toward Legitimacy, Wired News (Nov. 24, 2004), http://www.wired.com/news/ digiwood/0,1412,65836,00.html; Madden & Rainie, March 2005 Data Memo, supra, at 6-8 (percentage of current downloaders who have used paid services rose from 24% to 43% in a year; number using free services fell from 58% to 41%).

[545 U.S. 965] Thus, lawful music downloading services—those that charge the customer for downloading music and pay royalties to the copyright holder—have continued to grow and to produce substantial revenue. See Brief for Internet Law Faculty as Amicus Curiae 5-20; Bruno, Digital Entertainment: Piracy Fight Shows Encouraging Signs (Mar. 5, 2005), available at LEXIS, News Library, Billboard File (in 2004, consumers worldwide purchased more than 10 times the number of digital tracks purchased in 2003; global digital music market of $330 million in 2004 expected to double in 2005); Press Release, Informa Telecoms & Media, Steady Download Growth Defies P2P (global digital revenues will likely exceed $3 billion in 2010); Ashton, [International Federation of the Phonographic Industry] Predicts Downloads Will Hit the Mainstream, Music Week, Jan. 29, 2005, p. 6 (legal music sites and portable MP3 players "are helping to transform the digital music market" into "an everyday consumer experience"). And more advanced types of non-music-oriented peer-to-peer networks have also started to develop, drawing in part on the lessons of Grokster.

Finally, as Sony recognized, the legislative option remains available. Courts are less well suited than Congress to the task of "accommodat[ing] fully the varied permutations of competing interests that are inevitably implicated by such new technology." Sony, 464 U. S., at 431; see, e. g., Audio Home Recording Act of 1992, 106 Stat. 4237 (adding 17 U. S. C., ch. 10); Protecting Innovation and Art While Preventing Piracy: Hearing before the Senate Committee on the Judiciary, 108th Cong., 2d Sess. (2004).

I do not know whether these developments and similar alternatives will prove sufficient, but I am reasonably certain that, given their existence, a strong demonstrated need for modifying Sony (or for interpreting Sony's standard more strictly) has not yet been shown. That fact, along with the added risks that modification (or strict interpretation) would impose upon technological innovation, leads me to the conclusion that we should maintain Sony, reading its standard as I [545 U.S. 966] have read it. As so read, it requires affirmance of the Ninth Circuit's determination of the relevant aspects of the Sony question.

* * *

For these reasons, I disagree with JUSTICE GINSBURG, but I agree with the Court and join its opinion.

Notes:

[*] Briefs of amici curiae urging reversal were filed for the State of Utah et al. by Mark Shurtleff, Attorney General of Utah, and by the Attorneys General for their respective jurisdictions as follows: Troy King of Alabama, Gregg Renkes of Alaska, Terry Goddard of Arizona, Mike Beebe of Arkansas, M. Jane Brady of Delaware, Charles J. Crist, Jr., of Florida, Thurbert E. Baker of Georgia, Douglas B. Moylan of Guam, Mark J. Bennett of Hawaii, Lawrence G. Wasden of Idaho, Lisa Madigan of Illinois, Steve Carter of Indiana, Phill Kline of Kansas, Gregory D. Stumbo of Kentucky, Charles C. Foti, Jr., of Louisiana, Thomas F. Reilly of Massachusetts, Michael A. Cox of Michigan, Mike Hatch of Minnesota, Jim Hood of Mississippi, Jeremiah W. (Jay) Nixon of Missouri, Mike McGrath of Montana, Jon Bruning of Nebraska, Brian Sandoval of Nevada, Peter C. Harvey of New Jersey, Patricia A. Madrid of New Mexico, Roy Cooper of North Carolina, Wayne Stenehjem of North Dakota, Jim Petro of Ohio, W. A. Drew Edmondson of Oklahoma, Thomas W. Corbett, Jr., of Pennsylvania, Patrick Lynch of Rhode Island, Henry McMaster of South Carolina, Lawrence E. Long of South Dakota, Paul G. Summers of Tennessee, Greg Abbott of Texas, William H. Sorrell of Vermont, Jerry Kilgore of Virginia, Darrell V. McGraw, Jr., of West Virginia, and Peg Lautenschlager of Wisconsin; for the American Federation of Musicians of the United States and Canada et al. by George H. Cohen, Patricia Polach, and Laurence Gold; for the American Society of Composers, Authors and Publishers et al. by I. Fred Koenigsberg, Michael E. Salzman, and Marvin L. Berenson; for Americans for Tax Reform by Carter G. Phillips, Alan Charles Raul, Jay T. Jorgensen, and Eric A. Shumsky; for the Commissioner of Baseball et al. by Robert Alan Garrett and Hadrian R. Katz; for Defenders of Property Rights by Theodore B. Olson, Thomas H. Dupree, Jr., Matthew D. McGill, Nancie G. Marzulla, and Roger Marzulla; for International Rights Owners by Christopher Wolf; for Kids First Coalition et al. by Viet D. Dinh; for Law Professors et al. by James Gibson; for Macrovision Corp. by Geoffrey L. Beauchamp, Kelly G. Huller, and James H. Salter; for Napster, LLC, et al. by Barry I. Slotnick; for the National Academy of Recording Arts & Sciences, Inc., et al. by Jon A. Baumgarten and Jay L. Cooper; for the National Association of Broadcasters by Marsha J. MacBride, Jane E. Mago, Benjamin F. P. Ivins, and Jerianne Timmerman; for the National Association of Recording Merchandisers by Alan R. Malasky and Melanie Martin-Jones; for the Progress & Freedom Foundation by James V. DeLong; for the Video Software Dealers Association by John T. Mitchell; and for Professor Peter S. Menell et al. by Mr. Menell, pro se.

Briefs of amici curiae urging affirmance were filed for Altnet, Inc., by Roderick G. Dorman; for the American Civil Liberties Union et al. by Christopher A. Hansen, Steven R. Shapiro, Sharon M. McGowan, Ann Brick, and Jordan C. Budd; for the American Conservative Union et al. by David Post; for the Cellular Telecommunications & Internet Association et al. by Andrew G. McBride, Joshua S. Turner, Michael Altschul, James W. Olson, Frank L. Politano, Laura Kaster, Jeffrey A. Rackow, Grier C. Raclin, Michael Standard, John Thorne, Sarah B. Deutsch, and Paul J. Larkin, Jr.; for the Consumer Electronics Association et al. by Bruce G. Joseph and Scott E. Bain; for the Consumer Federation of America et al. by Peter Jaszi; for the Distributed Computing Industry Association by Mr. Dorman; for the Eagle Forum Education & Legal Defense Fund by Andrew L. Schlafly and Karen B. Tripp; for the Free Software Foundation et al. by Eben Moglen; for Intel Corp. by James M. Burger and Jonathan D. Hart; for Internet Law Faculty by William W. Fisher III and Jonathan Zittrain; for Law Professors by J. Glynn Lunney, Jr.; for the National Association of Shareholder and Consumer Attorneys by Kevin P. Roddy and Matthew E. Van Tine; for Sixty Intellectual Property and Technology Law Professors et al. by Deirdre K. Mulligan and Pamela Samuelson; for Sovereign Artists et al. by James R. Wheaton; for Computer Science Professor Harold Abelson et al. by James S. Tyre; for Professor Edward Lee et al. by Mr. Lee, pro se; for Charles Nesson by Mr. Nesson, pro se; and for Malla Pollack et al. by Ms. Pollack, pro se.

Briefs of amici curiae were filed for the American Intellectual Property Law Association by Rick D. Nydegger and Melvin C. Garner; for Audible Magic Corp. et al. by Bruce V. Spiva and Jeremy H. Stern; for Bridgemar Services, Ltd. d/b/a iMesh.com by Jeffrey A. Kimmel; for the Business Software Alliance by E. Edward Bruce and Robert A. Long, Jr.; for Creative Commons by Lawrence Lessig; for the Digital Media Association et al. by Lawrence Robbins, Alan Untereiner, Markham C. Erickson, and Jerry Berman; for Emerging Technology Companies by Michael Traynor and Matthew D. Brown; for IEEE-USA by Matthew J. Conigliaro, Andrew C. Greenberg, Joseph H. Lang, Jr., and Daniel E. Fisher; for Innovation Scholars and Economists by Laurence F. Pulgram; for the Intellectual Property Owners Association by James H. Pooley; for Media Studies Professors by Roy I. Liebman; for the National Venture Capital Association by Michael K. Kellogg, Mark L. Evans, and David L. Schwarz; for Sharman Networks Limited by Mr. Dorman; for SNOCAP, Inc., by Joel W. Nomkin; for Kenneth J. Arrow et al. by David A. Strauss; for Lee A. Hollaar by Lloyd W. Sadler; for U. S. Senator Patrick Leahy et al. by Mr. Leahy, pro se, and Senator Orrin G. Hatch, pro se; and for Felix Oberholzer-Gee et al. by Carl H. Settlemyer III and Arnold P. Lutzker.

[1] Peer-to-peer networks have disadvantages as well. Searches on peer-to-peer networks may not reach and uncover all available files because search requests may not be transmitted to every computer on the network. There may be redundant copies of popular files. The creator of the software has no incentive to minimize storage or bandwidth consumption, the costs of which are borne by every user of the network. Most relevant here, it is more difficult to control the content of files available for retrieval and the behavior of users.

[2] The studios and recording companies and the songwriters and music publishers filed separate suits against the defendants that were consolidated by the District Court.

[3] Subsequent versions of Morpheus, released after the record was made in this case, apparently rely not on Gnutella but on a technology called Neonet. These developments are not before us.

[4] There is some evidence that both Grokster and StreamCast previously operated supernodes, which compiled indexes of files available on all of the nodes connected to them. This evidence, pertaining to previous versions of the defendants' software, is not before us and would not affect our conclusions in any event.

[5] By comparison, evidence introduced by the plaintiffs in A&M Records, Inc. v. Napster, Inc., 239 F. 3d 1004 (CA9 2001), showed that 87% of files available on the Napster file-sharing network were copyrighted, id., at 1013.

[6] The Grokster founder contends that in answering these e-mails he often did not read them fully. App. 77, 769.

[7] The record makes clear that StreamCast developed these promotional materials but not whether it released them to the public. Even if these advertisements were not released to the public and do not show encouragement to infringe, they illuminate StreamCast's purposes.

[8] The mutual exclusivity of these values should not be overstated, however. On the one hand technological innovators, including those writing file-sharing computer programs, may wish for effective copyright protections for their work. See, e. g., Wu, When Code Isn't Law, 89 Va. L. Rev. 679, 750 (2003). (StreamCast itself was urged by an associate to "get [its] technology written down and [its intellectual property] protected." App. 866.) On the other hand the widespread distribution of creative works through improved technologies may enable the synthesis of new works or generate audiences for emerging artists. See Eldred v. Ashcroft, 537 U. S. 186, 223-226 (2003) (Stevens, J., dissenting); Van Houweling, Distributive Values in Copyright, 83 Texas L. Rev. 1535, 1539-1540, 1562-1564 (2005); Brief for Sovereign Artists et al. as Amici Curiae 11.

[9] We stated in Sony Corp. of America v. Universal City Studios, Inc., 464 U. S. 417 (1984), that "`the lines between direct infringement, contributory infringement and vicarious liability are not clearly drawn' . . . . [R]easoned analysis of [the Sony plaintiffs' contributory infringement claim] necessarily entails consideration of arguments and case law which may also be forwarded under the other labels, and indeed the parties . . . rely upon such arguments and authority in support of their respective positions on the issue of contributory infringement," id., at 435, n. 17 (quoting Universal City Studios, Inc. v. Sony Corp. of America, 480 F. Supp. 429, 457-458 (CD Cal. 1979)). In the present case MGM has argued a vicarious liability theory, which allows imposition of liability when the defendant profits directly from the infringement and has a right and ability to supervise the direct infringer, even if the defendant initially lacks knowledge of the infringement. See, e. g., Shapiro, Bernstein & Co. v. H. L. Green Co., 316 F. 2d 304, 308 (CA2 1963); Dreamland Ball Room, Inc. v. Shapiro, Bernstein & Co., 36 F. 2d 354, 355 (CA7 1929). Because we resolve the case based on an inducement theory, there is no need to analyze separately MGM's vicarious liability theory.

[10] Nor does the Patent Act's exemption from liability for those who distribute a staple article of commerce, 35 U. S. C. § 271(c), extend to those who induce patent infringement, § 271(b).

[11] Inducement has been codified in patent law. Ibid.

[12] Of course, in the absence of other evidence of intent, a court would be unable to find contributory infringement liability merely based on a failure to take affirmative steps to prevent infringement, if the device otherwise was capable of substantial noninfringing uses. Such a holding would tread too close to the Sony safe harbor.

[13] Grokster and StreamCast contend that any theory of liability based on their conduct is not properly before this Court because the rulings in the trial and appellate courts dealt only with the present versions of their software, not "past acts . . . that allegedly encouraged infringement or assisted . . . known acts of infringement." Brief for Respondents 14; see also id., at 34. This contention misapprehends the basis for their potential liability. It is not only that encouraging a particular consumer to infringe a copyright can give rise to secondary liability for the infringement that results. Inducement liability goes beyond that, and the distribution of a product can itself give rise to liability where evidence shows that the distributor intended and encouraged the product to be used to infringe. In such a case, the culpable act is not merely the encouragement of infringement but also the distribution of the tool intended for infringing use. See Kalem Co. v. Harper Brothers, 222 U. S. 55, 62-63 (1911); Cable/Home Communication Corp. v. Network Productions, Inc., 902 F. 2d 829, 846 (CA11 1990); A&M Records, Inc. v. Abdallah, 948 F. Supp. 1449, 1456 (CD Cal. 1996).

---------------

[14] JUSTICE BREYER finds in Sony Corp. of America v. Universal City Studios, Inc., 464 U. S. 417 (1984), a "clear" rule permitting contributory liability for copyright infringement based on distribution of a product only when the product "will be used almost exclusively to infringe copyrights." Post, at 957. But cf. Sony, 464 U. S., at 442 (recognizing "copyright holder's legitimate demand for effective—not merely symbolic—protection"). Sony, as I read it, contains no clear, near-exclusivity test. Nor have Courts of Appeals unanimously recognized JUSTICE BREYER'S clear rule. Compare A&M Records, Inc. v. Napster, Inc., 239 F. 3d 1004, 1021 (CA9 2001) ("[E]vidence of actual knowledge of specific acts of infringement is required to hold a computer system operator liable for contributory copyright infringement."), with In re Aimster Copyright Litigation, 334 F. 3d 643, 649-650 (CA7 2003) ("[W]hen a supplier is offering a product or service that has noninfringing as well as infringing uses, some estimate of the respective magnitudes of these uses is necessary for a finding of contributory infringement. . . . But the balancing of costs and benefits is necessary only in a case in which substantial noninfringing uses, present or prospective, are demonstrated."). See also Matthew Bender & Co. v. West Pub. Co., 158 F. 3d 693, 707 (CA2 1998) ("The Supreme Court applied [the Sony] test to prevent copyright holders from leveraging the copyrights in their original work to control distribution of . . . products that might be used incidentally for infringement, but that had substantial noninfringing uses. . . . The same rationale applies here [to products] that have substantial, predominant and noninfringing uses as tools for research and citation."). All Members of the Court agree, moreover, that "the Court of Appeals misapplied Sony," at least to the extent it read that decision to limit "secondary liability" to a hardly ever category, "quite beyond the circumstances to which the case applied." Ante, at 933.

[15] Grokster and StreamCast, in the Court of Appeals' view, would be entitled to summary judgment unless MGM could show that the software companies had knowledge of specific acts of infringement and failed to act on that knowledge—a standard the court held MGM could not meet. 380 F. 3d, at 1162-1163.

[16] JUSTICE BREYER finds support for summary judgment in this motley collection of declarations and in a survey conducted by an expert retained by MGM. Post, at 952-955. That survey identified 75% of the files available through Grokster as copyrighted works owned or controlled by the plaintiffs, and 15% of the files as works likely copyrighted. App. 439. As to the remaining 10% of the files, "there was not enough information to form reasonable conclusions either as to what those files even consisted of, and/or whether they were infringing or non-infringing." Id., at 479. Even assuming, as JUSTICE BREYER does, that the Sony Court would have absolved Sony of contributory liability solely on the basis of the use of the Betamax for authorized time-shifting, post, at 950-951, summary judgment is not inevitably appropriate here. Sony stressed that the plaintiffs there owned "well below 10%" of copyrighted television programming, 464 U. S., at 443, and found, based on trial testimony from representatives of the four major sports leagues and other individuals authorized to consent to home recording of their copyrighted broadcasts, that a similar percentage of program copying was authorized, id., at 424. Here, the plaintiffs allegedly control copyrights for 70% or 75% of the material exchanged through the Grokster and StreamCast software, 380 F. 3d 1154, 1158 (CA9 2004); App. 439, and the District Court does not appear to have relied on comparable testimony about authorized copying from copyright holders.

[17] The District Court's conclusion that "[p]laintiffs do not dispute that [d]efendants' software is being used, and could be used, for substantial noninfringing purposes," 259 F. Supp. 2d 1029, 1036 (CD Cal. 2003); accord 380 F. 3d, at 1161, is, to say the least, dubious. In the courts below and in this Court, MGM has continuously disputed any such conclusion. Brief for Motion Picture Studio and Recording Company Petitioners 30-38; Brief for MGM Plaintiffs-Appellants in No. 03-55894 etc. (CA9), p. 41; App. 356-357, 361-365.

Page 327

Page 328

        ARGUED: John Saul Edwards, Law Offices of John S. Edwards, Roanoke, VA; Leo Kayser, III, Kayser & Redfern, New York City, for Appellant. Patrick Joseph Carome, Wilmer, Cutler & Pickering, Washington, DC, for Appellee. ON BRIEF: John Payton, Samir Jain, Wilmer, Cutler & Pickering, Washington, DC; Randall J. Boe, America Online, Inc., Dulles, VA, for Appellee.

        Before WILKINSON, Chief Judge, RUSSELL, Circuit Judge, and BOYLE, Chief United States District Judge for the Eastern District of North Carolina, sitting by designation.

        Affirmed by published opinion. Chief Judge WILKINSON wrote the opinion, in which Judge RUSSELL and Chief Judge BOYLE joined.

        WILKINSON, Chief Judge:

        Kenneth Zeran brought this action against America Online, Inc. ("AOL"), arguing that AOL unreasonably delayed in removing defamatory messages posted by an unidentified third party, refused to post retractions of those messages, and failed to screen for similar postings thereafter. The district court granted judgment for AOL on the grounds that the Communications Decency Act of 1996 ("CDA")--47 U.S.C. § 230--bars Zeran's claims. Zeran appeals, arguing that § 230 leaves intact liability for interactive computer service providers who possess notice of defamatory material posted through their services. He also contends that § 230 does not apply here because his claims arise from AOL's alleged negligence prior to the CDA's enactment. Section 230, however, plainly immunizes computer service providers like AOL from liability for information that originates with third parties. Furthermore, Congress clearly expressed its intent that § 230 apply to lawsuits, like Zeran's, instituted after the CDA's enactment. Accordingly, we affirm the judgment of the district court.

I.

        "The Internet is an international network of interconnected computers," currently used by approximately 40 million people worldwide. Reno v. ACLU, --- U.S. ----, ----, 117 S.Ct. 2329, 2334, 138 L.Ed.2d 874 (1997). One of the many means by which individuals access the Internet is through an interactive computer service. These services offer not only a connection to the Internet as a whole,

Page 329

        The instant case comes before us on a motion for judgment on the pleadings, see Fed.R.Civ.P. 12(c), so we accept the facts alleged in the complaint as true. Bruce v. Riddle, 631 F.2d 272, 273 (4th Cir.1980). On April 25, 1995, an unidentified person posted a message on an AOL bulletin board advertising "Naughty Oklahoma T-Shirts." The posting described the sale of shirts featuring offensive and tasteless slogans related to the April 19, 1995, bombing of the Alfred P. Murrah Federal Building in Oklahoma City. Those interested in purchasing the shirts were instructed to call "Ken" at Zeran's home phone number in Seattle, Washington. As a result of this anonymously perpetrated prank, Zeran received a high volume of calls, comprised primarily of angry and derogatory messages, but also including death threats. Zeran could not change his phone number because he relied on its availability to the public in running his business out of his home. Later that day, Zeran called AOL and informed a company representative of his predicament. The employee assured Zeran that the posting would be removed from AOL's bulletin board but explained that as a matter of policy AOL would not post a retraction. The parties dispute the date that AOL removed this original posting from its bulletin board.

        On April 26, the next day, an unknown person posted another message advertising additional shirts with new tasteless slogans related to the Oklahoma City bombing. Again, interested buyers were told to call Zeran's phone number, to ask for "Ken," and to "please call back if busy" due to high demand. The angry, threatening phone calls intensified. Over the next four days, an unidentified party continued to post messages on AOL's bulletin board, advertising additional items including bumper stickers and key chains with still more offensive slogans. During this time period, Zeran called AOL repeatedly and was told by company representatives that the individual account from which the messages were posted would soon be closed. Zeran also reported his case to Seattle FBI agents. By April 30, Zeran was receiving an abusive phone call approximately every two minutes.

        Meanwhile, an announcer for Oklahoma City radio station KRXO received a copy of the first AOL posting. On May 1, the announcer related the message's contents on the air, attributed them to "Ken" at Zeran's phone number, and urged the listening audience to call the number. After this radio broadcast, Zeran was inundated with death threats and other violent calls from Oklahoma City residents. Over the next few days, Zeran talked to both KRXO and AOL representatives. He also spoke to his local police, who subsequently surveilled his home to protect his safety. By May 14, after an Oklahoma City newspaper published a story exposing the shirt advertisements as a hoax and after KRXO made an on-air apology, the number of calls to Zeran's residence finally subsided to fifteen per day.

        Zeran first filed suit on January 4, 1996, against radio station KRXO in the United States District Court for the Western District of Oklahoma. On April 23, 1996, he filed this separate suit against AOL in the same court. Zeran did not bring any action against the party who posted the offensive messages. ¹ After Zeran's suit against AOL was transferred to the Eastern District of Virginia pursuant to 28 U.S.C. § 1404(a), AOL answered Zeran's complaint and interposed 47 U.S.C. § 230 as an affirmative defense. AOL then moved for judgment on the pleadings pursuant to Fed.R.Civ.P. 12(c).

Page 330

II.

A.

        Because § 230 was successfully advanced by AOL in the district court as a defense to Zeran's claims, we shall briefly examine its operation here. Zeran seeks to hold AOL liable for defamatory speech initiated by a third party. He argued to the district court that once he notified AOL of the unidentified third party's hoax, AOL had a duty to remove the defamatory posting promptly, to notify its subscribers of the message's false nature, and to effectively screen future defamatory material. Section 230 entered this litigation as an affirmative defense pled by AOL. The company claimed that Congress immunized interactive computer service providers from claims based on information posted by a third party.

        The relevant portion of § 230 states: "No provider or user of an interactive computer service shall be treated as the publisher or speaker of any information provided by another information content provider." 47 U.S.C. § 230(c)(1). ² By its plain language, § 230 creates a federal immunity to any cause of action that would make service providers liable for information originating with a third-party user of the service. Specifically, § 230 precludes courts from entertaining claims that would place a computer service provider in a publisher's role. Thus, lawsuits seeking to hold a service provider liable for its exercise of a publisher's traditional editorial functions--such as deciding whether to publish, withdraw, postpone or alter content--are barred.

        The purpose of this statutory immunity is not difficult to discern. Congress recognized the threat that tort-based lawsuits pose to freedom of speech in the new and burgeoning Internet medium. The imposition of tort liability on service providers for the communications of others represented, for Congress, simply another form of intrusive government regulation of speech. Section 230 was enacted, in part, to maintain the robust nature of Internet communication and, accordingly, to keep government interference in the medium to a minimum. In specific statutory findings, Congress recognized the Internet and interactive computer services as offering "a forum for a true diversity of political discourse, unique opportunities for cultural development, and myriad avenues for intellectual activity." Id. § 230(a)(3). It also found that the Internet and interactive computer services "have flourished, to the benefit of all Americans, with a minimum of government regulation." Id. § 230(a)(4) (emphasis added). Congress further stated that it is "the policy of the United States ... to preserve the vibrant and competitive free market that presently exists for the Internet and other interactive computer services, unfettered by Federal or State regulation." Id. § 230(b)(2) (emphasis added).

        None of this means, of course, that the original culpable party who posts defamatory messages would escape accountability. While Congress acted to keep government regulation of the Internet to a minimum, it also found it to be the policy of the United States "to ensure vigorous enforcement of Federal criminal laws to deter and punish trafficking in obscenity, stalking, and harassment by means of computer." Id. § 230(b)(5). Congress made a policy choice, however, not to deter harmful online speech through the separate route of imposing tort liability on companies that serve as intermediaries

Page 331

        Congress' purpose in providing the § 230 immunity was thus evident. Interactive computer services have millions of users. See Reno v. ACLU, --- U.S. at ----, 117 S.Ct. at 2334 (noting that at time of district court trial, "commercial online services had almost 12 million individual subscribers"). The amount of information communicated via interactive computer services is therefore staggering. The specter of tort liability in an area of such prolific speech would have an obvious chilling effect. It would be impossible for service providers to screen each of their millions of postings for possible problems. Faced with potential liability for each message republished by their services, interactive computer service providers might choose to severely restrict the number and type of messages posted. Congress considered the weight of the speech interests implicated and chose to immunize service providers to avoid any such restrictive effect.

        Another important purpose of § 230 was to encourage service providers to self-regulate the dissemination of offensive material over their services. In this respect, § 230 responded to a New York state court decision, Stratton Oakmont, Inc. v. Prodigy Servs. Co., 1995 WL 323710 (N.Y.Sup.Ct. May 24, 1995). There, the plaintiffs sued Prodigy--an interactive computer service like AOL--for defamatory comments made by an unidentified party on one of Prodigy's bulletin boards. The court held Prodigy to the strict liability standard normally applied to original publishers of defamatory statements, rejecting Prodigy's claims that it should be held only to the lower "knowledge" standard usually reserved for distributors. The court reasoned that Prodigy acted more like an original publisher than a distributor both because it advertised its practice of controlling content on its service and because it actively screened and edited messages posted on its bulletin boards.

        Congress enacted § 230 to remove the disincentives to selfregulation created by the Stratton Oakmont decision. Under that court's holding, computer service providers who regulated the dissemination of offensive material on their services risked subjecting themselves to liability, because such regulation cast the service provider in the role of a publisher. Fearing that the specter of liability would therefore deter service providers from blocking and screening offensive material, Congress enacted § 230's broad immunity "to remove disincentives for the development and utilization of blocking and filtering technologies that empower parents to restrict their children's access to objectionable or inappropriate online material." 47 U.S.C. § 230(b)(4). In line with this purpose, § 230 forbids the imposition of publisher liability on a service provider for the exercise of its editorial and self-regulatory functions.

B.

        Zeran argues, however, that the § 230 immunity eliminates only publisher liability, leaving distributor liability intact. Publishers can be held liable for defamatory statements contained in their works even absent proof that they had specific knowledge of the statement's inclusion. W. Page Keeton et al., Prosser and Keeton on the Law of Torts § 113, at 810 (5th ed.1984). According to Zeran, interactive computer service providers like AOL are normally considered instead to be distributors, like traditional news vendors or book sellers. Distributors cannot be held liable for defamatory statements contained in the materials they distribute unless it is proven at a minimum that they have actual knowledge of the defamatory statements upon which liability is predicated. Id. at 811 (explaining that distributors are not liable "in the absence of proof that they knew or had reason to know of the existence of defamatory matter contained in matter published"). Zeran contends that he provided AOL with sufficient notice of the defamatory statements appearing on the company's bulletin board. This notice is significant, says Zeran, because AOL could be held liable as a distributor only if it acquired knowledge of the defamatory statements' existence.

        Because of the difference between these two forms of liability, Zeran contends that the term "distributor" carries a legally distinct meaning from the term "publisher."

Page 332

        The terms "publisher" and "distributor" derive their legal significance from the context of defamation law. Although Zeran attempts to artfully plead his claims as ones of negligence, they are indistinguishable from a garden variety defamation action. Because the publication of a statement is a necessary element in a defamation action, only one who publishes can be subject to this form of tort liability. Restatement (Second) of Torts § 558(b) (1977); Keeton et al., supra, § 113, at 802. Publication does not only describe the choice by an author to include certain information. In addition, both the negligent communication of a defamatory statement and the failure to remove such a statement when first communicated by another party--each alleged by Zeran here under a negligence label--constitute publication. Restatement (Second) of Torts § 577; see also Tacket v. General Motors Corp., 836 F.2d 1042, 1046-47 (7th Cir.1987). In fact, every repetition of a defamatory statement is considered a publication. Keeton et al., supra, § 113, at 799.

        In this case, AOL is legally considered to be a publisher. "[E]very one who takes part in the publication ... is charged with publication." Id. Even distributors are considered to be publishers for purposes of defamation law:

Those who are in the business of making their facilities available to disseminate the writings composed, the speeches made, and the information gathered by others may also be regarded as participating to such an extent in making the books, newspapers, magazines, and information available to others as to be regarded as publishers. They are intentionally making the contents available to others, sometimes without knowing all of the contents--including the defamatory content--and sometimes without any opportunity to ascertain, in advance, that any defamatory matter was to be included in the matter published.

        Id. at 803. AOL falls squarely within this traditional definition of a publisher and, therefore, is clearly protected by § 230's immunity.

        Zeran contends that decisions like Stratton Oakmont and Cubby, Inc. v. CompuServe Inc., 776 F.Supp. 135 (S.D.N.Y.1991), recognize a legal distinction between publishers and distributors. He misapprehends, however, the significance of that distinction for the legal issue we consider here. It is undoubtedly true that mere conduits, or distributors, are subject to a different standard of liability. As explained above, distributors must at a minimum have knowledge of the existence of a defamatory statement as a prerequisite to liability. But this distinction signifies only that different standards of liability may be applied within the larger publisher category, depending on the specific type of publisher concerned. See Keeton et al., supra, § 113, at 799-800 (explaining that every party involved is charged with publication, although degrees of legal responsibility differ). To the extent that decisions like Stratton and Cubby utilize the terms "publisher" and "distributor" separately, the decisions correctly describe two different standards of liability. Stratton and Cubby do not, however, suggest that distributors are not also a type of publisher for purposes of defamation law.

        Zeran simply attaches too much importance to the presence of the distinct notice element in distributor liability. The simple fact of notice surely cannot transform one from an original publisher to a distributor in the eyes of the law. To the contrary, once a computer service provider receives notice of a potentially defamatory posting, it is thrust into the role of a traditional publisher. The computer service provider must decide whether to publish, edit, or withdraw the posting. In this respect, Zeran seeks to impose liability on AOL for assuming the

Page 333

        Our view that Zeran's complaint treats AOL as a publisher is reinforced because AOL is cast in the same position as the party who originally posted the offensive messages. According to Zeran's logic, AOL is legally at fault because it communicated to third parties an allegedly defamatory statement. This is precisely the theory under which the original poster of the offensive messages would be found liable. If the original party is considered a publisher of the offensive messages, Zeran certainly cannot attach liability to AOL under the same theory without conceding that AOL too must be treated as a publisher of the statements.

        Zeran next contends that interpreting § 230 to impose liability on service providers with knowledge of defamatory content on their services is consistent with the statutory purposes outlined in Part IIA. Zeran fails, however, to understand the practical implications of notice liability in the interactive computer service context. Liability upon notice would defeat the dual purposes advanced by § 230 of the CDA. Like the strict liability imposed by the Stratton Oakmont court, liability upon notice reinforces service providers' incentives to restrict speech and abstain from self-regulation.

        If computer service providers were subject to distributor liability, they would face potential liability each time they receive notice of a potentially defamatory statement--from any party, concerning any message. Each notification would require a careful yet rapid investigation of the circumstances surrounding the posted information, a legal judgment concerning the information's defamatory character, and an on-the-spot editorial decision whether to risk liability by allowing the continued publication of that information. Although this might be feasible for the traditional print publisher, the sheer number of postings on interactive computer services would create an impossible burden in the Internet context. Cf. Auvil v. CBS 60 Minutes, 800 F.Supp. 928, 931 (E.D.Wash.1992) (recognizing that it is unrealistic for network affiliates to "monitor incoming transmissions and exercise on-the-spot discretionary calls"). Because service providers would be subject to liability only for the publication of information, and not for its removal, they would have a natural incentive simply to remove messages upon notification, whether the contents were defamatory or not. See Philadelphia Newspapers, Inc. v. Hepps, 475 U.S. 767, 777, 106 S.Ct. 1558, 1564, 89 L.Ed.2d 783 (1986) (recognizing that fears of unjustified liability produce a chilling effect antithetical to First Amendment's protection of speech). Thus, like strict liability, liability upon notice has a chilling effect on the freedom of Internet speech.

        Similarly, notice-based liability would deter service providers from regulating the dissemination of offensive material over their own services. Any efforts by a service provider to investigate and screen material posted on its service would only lead to notice of potentially defamatory material more frequently and thereby create a stronger basis for liability. Instead of subjecting themselves to further possible lawsuits, service providers would likely eschew any attempts at self-regulation.

        More generally, notice-based liability for interactive computer service providers would provide third parties with a no-cost means to create the basis for future lawsuits. Whenever one was displeased with the speech of another party conducted over an interactive computer service, the offended party could simply "notify" the relevant service provider, claiming the information to be legally defamatory. In light of the vast amount of speech communicated through interactive computer services, these notices could produce an impossible burden for service providers, who would be faced with ceaseless choices of suppressing controversial speech or sustaining prohibitive liability. Because the probable effects of distributor liability on the vigor of Internet speech and on service provider self-regulation are directly contrary to § 230's statutory purposes, we will not assume that Congress intended to leave liability upon notice intact.

        Zeran finally contends that the interpretive canon favoring retention of common law principles unless Congress speaks directly to the issue counsels a restrictive reading of the

Page 334

        The decision cited by Zeran, United States v. Texas, also recognized that abrogation of common law principles is appropriate when a contrary statutory purpose is evident. Id. This is consistent with the Court's earlier cautions against courts' application of the canon with excessive zeal: " 'The rule that statutes in derogation of the common law are to be strictly construed does not require such an adherence to the letter as would defeat an obvious legislative purpose or lessen the scope plainly intended to be given to the measure.' " Isbrandtsen Co. v. Johnson, 343 U.S. 779, 783, 72 S.Ct. 1011, 1014, 96 L.Ed. 1294 (1952) (quoting Jamison v. Encarnacion, 281 U.S. 635, 640, 50 S.Ct. 440, 442, 74 L.Ed. 1082 (1930)); cf. Astoria Fed. Sav. & Loan Ass'n v. Solimino, 501 U.S. 104, 110-11, 111 S.Ct. 2166, 2170-71, 115 L.Ed.2d 96 (1991) (statute need not expressly delimit manner in which common law principle is abrogated). Zeran's argument flies in the face of this warning. As explained above, interpreting § 230 to leave distributor liability in effect would defeat the two primary purposes of the statute and would certainly "lessen the scope plainly intended" by Congress' use of the term "publisher."

        Section 230 represents the approach of Congress to a problem of national and international dimension. The Supreme Court underscored this point in Reno v. ACLU, finding that the Internet allows "tens of millions of people to communicate with one another and to access vast amounts of information from around the world. [It] is 'a unique and wholly new medium of worldwide human communication.' " --- U.S. at ----, 117 S.Ct. at 2334 (citation omitted). Application of the canon invoked by Zeran here would significantly lessen Congress' power, derived from the Commerce Clause, to act in a field whose international character is apparent. While Congress allowed for the enforcement of "any State law that is consistent with [ § 230]," 47 U.S.C. § 230(d)(3), it is equally plain that Congress' desire to promote unfettered speech on the Internet must supersede conflicting common law causes of action. Section 230(d)(3) continues: "No cause of action may be brought and no liability may be imposed under any State or local law that is inconsistent with this section." With respect to federal-state preemption, the Court has advised: "[W]hen Congress has 'unmistakably ... ordained,' that its enactments alone are to regulate a part of commerce, state laws regulating that aspect of commerce must fall. The result is compelled whether Congress' command is explicitly stated in the statute's language or implicitly contained in its structure and purpose." Jones v. Rath Packing Co., 430 U.S. 519, 525, 97 S.Ct. 1305, 1309, 51 L.Ed.2d 604 (1977) (citations omitted). Here, Congress' command is explicitly stated. Its exercise of its commerce power is clear and counteracts the caution counseled by the interpretive canon favoring retention of common law principles.

III.

        The CDA was signed into law and became effective on February 8, 1996. Zeran did not file his complaint until April 23, 1996. Zeran contends that even if § 230 does bar the type of claim he brings here, it cannot be applied retroactively to bar an action arising from AOL's alleged misconduct prior to the CDA's enactment. We disagree. Section 230 applies by its plain terms to complaints brought after the CDA became effective. As noted in Part IIB, the statute provides, in part: "No cause of action may be brought and no liability may be imposed under any State or local law that is inconsistent with this section." 47 U.S.C. § 230(d)(3).

        Initially, it is doubtful that a retroactivity issue is even presented here. Retroactivity concerns arise when a statute applies to conduct predating its enactment. Section 230 does not directly regulate the activities of interactive computer service providers like AOL. Instead, § 230 is addressed only to the bringing of a cause of action. Here, Zeran

Page 335

        Even if this were a case implicating the application of a federal statute to pre-enactment events, the Supreme Court's Landgraf framework would nevertheless require § 230's application to Zeran's claims. Landgraf instructs us first "to determine whether Congress has expressly prescribed the statute's proper reach." Landgraf v. USI Film Prods., 511 U.S. 244, 280, 114 S.Ct. 1483, 1505, 128 L.Ed.2d 229 (1994). This case can be resolved at this first step. In § 230(d)(3), Congress clearly expressed its intent that the statute apply to any complaint instituted after its effective date, regardless of when the relevant conduct giving rise to the claims occurred. Other circuits have interpreted similar statutory language to clearly express Congress' intent that the relevant statutes apply to bar new actions under statutorily specified conditions. See Wright v. Morris, 111 F.3d 414, 418 (6th Cir.1997) (holding language "No action shall be brought ...," 42 U.S.C. § 1997e(a), to "expressly govern[ ] the bringing of new actions"), cert. denied, --- U.S. ----, 118 S.Ct. 263, 139 L.Ed.2d 190 (1997); Abdul-Wadood v. Nathan, 91 F.3d 1023, 1025 (7th Cir.1996) (holding language "In no event shall a prisoner bring a civil action or appeal a judgment ...," 28 U.S.C. § 1915(g), to govern the bringing of new actions or filing of new appeals).

        If we were to find a directive as plain as § 230(d)(3) to be ambiguous as to Congress' intent, we would be announcing a new superclear-statement condition for the retroactive operation of statutes. Such a jurisprudential shift would be both unwise and contrary to the Court's admonitions in Landgraf: "Retroactivity provisions often serve entirely benign and legitimate purposes, whether to respond to emergencies, to correct mistakes, to prevent circumvention of a new statute in the interval immediately preceding its passage, or simply to give comprehensive effect to a new law Congress considers salutary." 511 U.S. at 267-68, 114 S.Ct. at 1498. Here, Congress decided that free speech on the Internet and self-regulation of offensive speech were so important that § 230 should be given immediate, comprehensive effect.

        There finally is a significant contrast between statutes that impose new liabilities for already-completed conduct and statutes that govern litigants' access to courts. For example, courts often apply intervening statutes that restrict a court's jurisdiction. See Landgraf, 511 U.S. at 274, 114 S.Ct. at 1501-02. Section 230 neither imposes any new liability on Zeran nor takes away any rights acquired under prior law. No person has a vested right in a nonfinal tort judgment, much less an unfiled tort claim. Hammond v. United States, 786 F.2d 8, 12 (1st Cir.1986). Furthermore, Zeran cannot point to any action he took in reliance on the law prior to § 230's enactment. Because § 230 has no untoward retroactive effect, even the presumption against statutory retroactivity absent an express directive from Congress is of no help to Zeran here.

IV.

        For the foregoing reasons, we affirm the judgment of the district court.

        AFFIRMED.

---------------

1 Zeran maintains that AOL made it impossible to identify the original party by failing to maintain adequate records of its users. The issue of AOL's record keeping practices, however, is not presented by this appeal.

2 Section 230 defines "interactive computer service" as "any information service, system, or access software provider that provides or enables computer access by multiple users to a computer server, including specifically a service or system that provides access to the Internet and such systems operated or services offered by libraries or educational institutions." 47 U.S.C. § 230(e)(2). The term"information content provider" is defined as "any person or entity that is responsible, in whole or in part, for the creation or development of information provided through the Internet or any other interactive computer service." Id. § 230(e)(3). The parties do not dispute that AOL falls within the CDA's "interactive computer service" definition and that the unidentified third party who posted the offensive messages here fits the definition of an "information content provider."

521 F.3d 1157 (2008)

Nos. 04-56916, 04-57173.

Argued and Submitted December 12, 2007.
Filed April 3, 2008.

[1158] [1159] [1160] Michael Evans, Christopher Brancart, Brancart & Brancart, Pescadero, CA; Gary Rhoades, Rhoades & Al-Mansour, Los Angeles, CA, for the plaintiffs-appellants/cross-appellees.

Timothy L. Alger, Kent J. Bullard, Steven B. Stiglitz and Lesley E. Williams, [1161] Quinn Emanuel Urquhart Oliver & Hedges, LLP, Los Angeles, CA, for the defendant-appellee/cross-appellant.

Kelli L. Sager, Los Angeles, CA; Thomas R. Burke, San Francisco, CA; Bruce E.H. Johnson and Ambika K. Doran, Davis Wright Tremaine LLP, Seattle, WA, for News Organizations as amici curiae in support of the defendant-appellee.

Ann Brick, Margaret C. Crosby and Nicole A. Ozer, American Civil Liberties Union Foundation of Northern California, San Francisco, CA, for American Civil Liberties Union of Northern California as amicus curiae in support of neither party.

John P. Relman, Stephen M. Dane and D. Scott Chang, Relman & Dane PLLC, Washington, DC; Joseph D. Rich and Nicole Birch, Lawyers' Committee for Civil Rights Under Law, Washington, DC, for National Fair Housing Alliance and Lawyers' Committee for Civil Rights Under Law as amici curiae in support of the plaintiffs-appellants.

Before: ALEX KOZINSKI, Chief Judge, STEPHEN REINHARDT, PAMELA ANN RYMER, BARRY G. SILVERMAN, M. MARGARET McKEOWN, W. FLETCHER, RAYMOND C. FISHER, RICHARD A. PAEZ, CARLOS T. BEA, MILAN D. SMITH, JR. and N. RANDY SMITH, Circuit Judges.

KOZINSKI, Chief Judge:

We plumb the depths of the immunity provided by section 230 of the Communications Decency Act of 1996 ("CDA").

Facts^[1]

Defendant Roommate.com, LLC ("Roommate") operates a website designed to match people renting out spare rooms with people looking for a place to live.^[2] At the time of the district court's disposition, Roommate's website featured approximately 150,000 active listings and received around a million page views a day. Roommate seeks to profit by collecting revenue from advertisers and subscribers.

Before subscribers can search listings or post^[3] housing opportunities on Roommate's website, they must create profiles, a process that requires them to answer a series of questions. In addition to requesting basic information — such as name, location and email address — Roommate requires each subscriber to disclose his sex, sexual orientation and whether he would bring children to a household. Each subscriber must also describe his preferences in roommates with respect to the same three criteria: sex, sexual orientation and whether they will bring children to the household. The site also encourages subscribers to provide "Additional Comments" describing themselves and their desired roommate in an open-ended essay. After a new subscriber completes the application, Roommate assembles his answers into a "profile page." The profile page [1162] displays the subscriber's pseudonym, his description and his preferences, as divulged through answers to Roommate's questions.

Subscribers can choose between two levels of service: Those using the site's free service level can create their own personal profile page, search the profiles of others and send personal email messages. They can also receive periodic emails from Roommate, informing them of available housing opportunities matching their preferences. Subscribers who pay a monthly fee also gain the ability to read emails from other users, and to view other subscribers' "Additional Comments."

The Fair Housing Councils of the San Fernando Valley and San Diego ("Councils") sued Roommate in federal court, alleging that Roommate's business violates the federal Fair Housing Act ("FHA"), 42 U.S.C. § 3601 et seq., and California housing discrimination laws.^[4] Councils claim that Roommate is effectively a housing broker doing online what it may not lawfully do off-line. The district court held that Roommate is immune under section 230 of the CDA, 47 U.S.C. § 230(c), and dismissed the federal claims without considering whether Roommate's actions violated the FHA. The court then declined to exercise supplemental jurisdiction over the state law claims. Councils appeal the dismissal of the FHA claim and Roommate cross-appeals the denial of attorneys' fees.

Analysis

Section 230 of the CDA^[5] immunizes providers of interactive computer services^[6] against liability arising from content created by third parties: "No provider . . . of an interactive computer service shall be treated as the publisher or speaker of any information provided by another information content provider." 47 U.S.C. § 230(c).^[7] This grant of immunity applies only if the interactive computer service provider is not also an "information content provider," which is defined as someone who is "responsible, in whole or in part, for the creation or development of the offending content. Id. § 230(f)(3).

A website operator can be both a service provider and a content provider: If it passively displays content that is created entirely by third parties, then it is only a service provider with respect to that content. But as to content that it creates itself, or is "responsible, in whole or in part" for creating or developing, the website is also a content provider. Thus, a website may be immune from liability for [1163] some of the content it displays to the public but be subject to liability for other content.^[8]

Section 230 was prompted by a state court case holding Prodigy^[9] responsible for a libelous message posted on one of its financial message boards.^[10]See Stratton Oakmont, Inc. v. Prodigy Servs. Co., 1995 WL 323710 (N.Y.Sup.Ct. May 24, 1995) (unpublished). The court there found that Prodigy had become a "publisher" under state law because it voluntarily deleted some messages from its message boards "on the basis of offensiveness and `bad taste,'" and was therefore legally responsible for the content of defamatory messages that it failed to delete. Id. at *4. The Stratton Oakmont court reasoned that Prodigy's decision to perform some voluntary self-policing made it akin to a newspaper publisher, and thus responsible for messages on its bulletin board that defamed third parties. The court distinguished Prodigy from CompuServe,^[11] which had been released from liability in a similar defamation case because CompuServe "had no opportunity to review the contents of the publication at issue before it was uploaded into CompuServe's computer banks." Id.; see Cubby, Inc. v. CompuServe, Inc., 776 F.Supp. 135, 140 (S.D.N.Y.1991). Under the reasoning of Stratton Oakmont, online service providers that voluntarily filter some messages become liable for all messages transmitted, whereas providers that bury their heads in the sand and ignore problematic posts altogether escape liability. Prodigy claimed that the "sheer volume" of message board postings it received — at the time, over 60,000 a day — made manual review of every message impossible; thus, if it were forced to choose between taking responsibility for all messages and deleting no messages at all, it would have to choose the latter course. Stratton Oakmont, 1995 WL 323710 at *3.

In passing section 230, Congress sought to spare interactive computer services this grim choice by allowing them to perform some editing on user-generated content without thereby becoming liable for all defamatory or otherwise unlawful messages that they didn't edit or delete. In other words, Congress sought to immunize the removal of user-generated content, not the creation of content: "[S]ection [230] provides `Good Samaritan' protections from civil liability for providers . . . of an interactive computer service for actions to restrict . . . access to objectionable online material. One of the specific purposes of this section is to overrule Stratton-Oakmont [sic] v. Prodigy and any other similar decisions which have treated such providers . . . as publishers or speakers of content that is not their own because they have restricted access to objectionable material." H.R.Rep. No. 104-458 (1996) (Conf.Rep.), as reprinted in 1996 U.S.C.C.A.N. 10 (emphasis added).<sup>[12]</sup> Indeed, the section is titled "Protection for `good Samaritan' blocking and [1164] screening of offensive material" and, as the Seventh Circuit recently held, the substance of section 230(c) can and should be interpreted consistent with its caption. Chicago Lawyers' Committee for Civil Rights Under Law, Inc. v. craigslist, Inc., 516 F.3d 666, ____ (7th Cir.2008) (quoting Doe v. GTE Corp., 347 F.3d 655, 659-60 (7th Cir.2003)).

With this backdrop in mind, we examine three specific functions performed by Roommate that are alleged to violate the Fair Housing Act and California law.

1. Councils first argue that the questions Roommate poses to prospective subscribers during the registration process violate the Fair Housing Act and the analogous California law. Councils allege that requiring subscribers to disclose their sex, family status and sexual orientation "indicates" an intent to discriminate against them, and thus runs afoul of both the FHA and state law.^[13]

Roommate created the questions and choice of answers, and designed its website registration process around them. Therefore, Roommate is undoubtedly the "information content provider" as to the questions and can claim no immunity for posting them on its website, or for forcing subscribers to answer them as a condition of using its services.

Here, we must determine whether Roommate has immunity under the CDA because Councils have at least a plausible claim that Roommate violated state and federal law by merely posing the questions. We need not decide whether any of Roommate's questions actually violate the Fair' Housing Act or California law, or whether they are protected by the First Amendment or other constitutional guarantees, see craigslist, at 1166-67; we leave those issues for the district court on remand. Rather, we examine the scope of plaintiffs' substantive claims only insofar as necessary to determine whether section 230 immunity applies. However, we note that asking questions certainly can violate the Fair Housing Act and analogous laws in the physical world.^[14] For example, a real estate broker may not inquire as to the race of a prospective buyer, and an employer may not inquire as to the religion of a prospective employee. If such questions are unlawful when posed face-to-face or by telephone, they don't magically become lawful when asked electronically online. The Communications Decency Act was not meant to create a lawless noman's-land on the Internet.^[15]

[1165] Councils also claim that requiring subscribers to answer the questions as a condition of using Roommate's services unlawfully "cause[s]" subscribers to make a "statement . . . with respect to the sale or rental of a dwelling that indicates [a] preference, limitation, or discrimination," in violation of 42 U.S.C. § 3604(c). The CDA does not grant immunity for inducing third parties to express illegal preferences. Roommate's own acts — posting the questionnaire and requiring answers to it — are entirely its doing and thus section 230 of the CDA does not apply to them. Roommate is entitled to no immunity.^[16]

2. Councils also charge that Roommate's development and display of subscribers' discriminatory preferences is unlawful. Roommate publishes a "profile page" for each subscriber on its website. The page describes the client's personal information — such as his sex, sexual orientation and whether he has children — as well as the attributes of the housing situation he seeks. The content of these pages is drawn directly from the registration process: For example, Roommate requires subscribers to specify, using a drop-down menu^[17] provided by Roommate, whether they are "Male" or "Female" and then displays that information on the profile page. Roommate also requires subscribers who are listing available housing to disclose whether there are currently "Straight male(s)," "Gay male(s)," "Straight female(s)" or "Lesbian(s)" living in the dwelling. Subscribers who are seeking housing must make a selection from a drop-down menu, again provided by Roommate, to indicate whether they are willing to live with "Straight or gay" males, only with "Straight" males, only with "Gay" males or, with "No males." Similarly, Roommate requires subscribers listing housing to disclose whether there are "Children present" or "Children not present" and requires housing seekers to say "I will live with children" or "I will not live with children." Roommate then displays these answers, along with other information, on the subscriber's profile page. This information is obviously included to help subscribers decide which housing opportunities to pursue and which to bypass. In addition, Roommate itself uses this information to channel subscribers away from listings where the individual offering housing has expressed preferences that aren't compatible with the subscriber's answers.

The dissent tilts at windmills when it shows, quite convincingly, that Roommate's subscribers are information content providers who create the profiles by picking among options and providing their own answers. Dissent at 1180-82. There is no disagreement on this point. But, the fact that users are information content providers does not preclude Roommate from also being an information content provider by helping "develop" at least "in part" the information in the profiles. As we explained in Batzel, the party responsible for putting information online may be subject to liability, even if the information originated with a user. See Batzel v. Smith, 333 F.3d 1018, 1033 (9th Cir.2003).^[18]

[1166] Here, the part of the profile that is alleged to offend the Fair Housing Act and state housing discrimination laws — the information about sex, family status and sexual orientation — is provided by subscribers in response to Roommate's questions, which they cannot refuse to answer if they want to use defendant's services. By requiring subscribers to provide the information as a condition of accessing its service, and by providing a limited set of pre-populated answers, Roommate becomes much more than a passive transmitter of information provided by others; it becomes the developer, at least in part, of that information. And section 230 provides immunity only if the interactive computer service does not "creat[e] or develop[]" the information "in whole or in part." See 47 U.S.C. § 230(f)(3).

Our dissenting colleague takes a much narrower view of what it means to "develop" information online, and concludes that Roommate does not develop the information because "[a]ll Roommate does is to provide a form with options for standardized answers." Dissent at 1182. But Roommate does much more than provide options. To begin with, it asks discriminatory questions that even the dissent grudgingly admits are not entitled to CDA immunity. Dissent at 1177 n. 5. The FHA makes it unlawful to ask certain discriminatory questions for a very good reason: Unlawful questions solicit (a.k.a."develop") unlawful answers. Not only does Roommate ask these questions, Roommate makes answering the discriminatory questions a condition of doing business. This is no different from a real estate broker in real life saying, "Tell me whether you're Jewish or you can find yourself another broker." When a business enterprise extracts such information from potential customers as a condition of accepting them as clients, it is no stretch to say that the enterprise is responsible, at least in part, for developing that information. For the dissent to claim that the information in such circumstances is "created solely by" the customer, and that the business has not helped in the least to develop it, Dissent at 1181-82, strains both credulity and English.^[19]

Roommate also argues that it is not responsible for the information on the profile page because it is each subscriber's action that leads to publication of his particular profile — in other words, the user pushes the last button or takes the last act before publication. We are not convinced that this is even true,^[20] but don't see why it matters anyway. The projectionist in the theater may push the last button before a film is displayed on the screen, but surely this doesn't make him the sole producer of [1167] the movie. By any reasonable use of the English language, Roommate is "responsible" at least "in part" for each subscriber's profile page, because every such page is a collaborative effort between Roommate and the subscriber.

Similarly, Roommate is not entitled to CDA immunity for the operation of its search system, which filters listings, or of its email notification system, which directs emails to subscribers according to discriminatory criteria.^[21] Roommate designed its search system so it would steer users based on the preferences and personal characteristics that Roommate itself forces subscribers to disclose. If Roommate has no immunity for asking the discriminatory questions, as we concluded above, see pp. 1163-65 supra, it can certainly have no immunity for using the answers to the unlawful questions to limit who has access to housing.

For example, a subscriber who self-identifies as a "Gay male" will not receive email notifications of new housing opportunities supplied by owners who limit the universe of acceptable tenants to "Straight male(s)," "Straight female(s)" and "Lesbian(s)." Similarly, subscribers with children will not be notified of new listings where the owner specifies "no children." Councils charge that limiting the information a subscriber can access based on that subscriber's protected status violates the Fair Housing Act and state housing discrimination laws. It is, Councils allege, no different from a real estate broker saying to a client: "Sorry, sir, but I can't show you any listings on this block because you are [gay/female/black/a parent]." If such screening is prohibited when practiced in person or by telephone, we see no reason why Congress would have wanted to make it lawful to profit from it online.

Roommate's search function is similarly designed to steer users based on discriminatory criteria. Roommate's search engine thus differs materially from generic search engines such as Google, Yahoo! and MSN Live Search, in that Roommate designed its system to use allegedly unlawful criteria so as to limit the results of each search, and to force users to participate in its discriminatory process. In other words, Councils allege that Roommate's search is designed to make it more difficult or impossible for individuals with certain protected characteristics to find housing — something the law prohibits. By contrast, ordinary search engines do not use unlawful criteria to limit the scope of searches conducted on them, nor are they designed to achieve illegal ends — as Roommate's search function is alleged to do here. Therefore, such search engines play no part in the "development" of any unlawful searches. See 47 U.S.C. § 230(f)(3).

It's true that the broadest sense of the term "develop" could include the functions of an ordinary search engine — indeed, just about any function performed by a website. But to read the term so broadly would defeat the purposes of section 230 by swallowing up every bit of the immunity that the section otherwise provides. At the same time, reading the exception for co-developers as applying only to content that originates entirely with the website — as the dissent would seem to suggest — ignores the words "development . . . in part" in the statutory passage "creation or development in whole or in part." 47 U.S.C. § 230(f)(3) (emphasis added). We believe that both the immunity for passive conduits and the exception for co-developers must be given their proper scope and, to that end, we interpret the term "development" as referring not merely [1168] to augmenting the content generally, but to materially contributing to its alleged unlawfulness. In other words, a website helps to develop unlawful content, and thus falls within the exception to section 230, if it contributes materially to the alleged illegality of the conduct.

The dissent accuses us of "rac[ing] past the plain language of the statute," dissent at 1185, but we clearly do pay close attention to the statutory language, particularly the word "develop;" which we spend many pages exploring. The dissent may disagree with our definition of the term, which is entirely fair, but surely our dissenting colleague is mistaken in suggesting we ignore the term. Nor is the statutory language quite as plain as the dissent would have it. Dissent at 1183-85. Quoting selectively from the dictionary, the dissent comes up with an exceedingly narrow definition of this rather complex and multi faceted term.^[22] Dissent at 1184 (defining development as "gradual advance or growth through progressive changes") (quoting Webster's Third New International Dictionary 618 (2002)). The dissent does not pause to consider how such a definition could apply to website content at all, as it excludes the kinds of swift and disorderly changes that are the hallmark of growth on the Internet. Had our dissenting colleague looked just a few lines lower on the same page of the same edition of the same dictionary, she would have found another definition of "development" that is far more suitable to the context in which we operate: "making usable or available." Id. The dissent does not explain why the definition it has chosen reflects the statute's "plain meaning," while the ones it bypasses do not.

More fundamentally, the dissent does nothing at all to grapple with the difficult statutory problem posed by the fact that section 230(c) uses both "create" and "develop" as separate bases for loss of immunity. Everything that the dissent includes within its cramped definition of "development" fits just as easily within the definition of "creation" — which renders the term "development" superfluous. The dissent makes no attempt to explain or offer examples as to how its interpretation of the statute leaves room for "development" as a separate basis for a website to lose its immunity, yet we are advised by the Supreme Court that we must give meaning to all statutory terms, avoiding redundancy or duplication wherever possible. See Park `N Fly, Inc. v. Dollar Park & Fly, Inc., 469 U.S. 189, 197, 105 S.Ct. 658, 83 L.Ed.2d 582 (1985).

While content to pluck the "plain meaning" of the statute from a dictionary definition that predates the Internet by decades, compare Webster's Third New International Dictionary 618 (1963) with Webster's Third New International Dictionary 618 (2002) (both containing "gradual advance or growth through progressive changes"), the dissent overlooks the far more relevant definition of "[web] content development" in Wikipedia: "the process of researching, writing, gathering, organizing and editing information for publication on web sites." Wikipedia, Content Development (Web), http://en.wikipedia. org/w/index.php?title=Content — development — web. & oldid=XXXXXXXXX (last visited Mar. 19, 2008). Our interpretation of "development" is entirely in line with the context-appropriate meaning of the term, [1169] and easily fits the activities Roommate engages in.

In an abundance of caution, and to avoid the kind of misunderstanding the dissent seems to encourage, we offer a few examples to elucidate what does and does not amount to "development" under section 230 of the Communications Decency Act: If an individual uses an ordinary search engine to query for a "white roommate," the search engine has not contributed to i any alleged unlawfulness in the individual's conduct; providing neutral tools to carry out what may be unlawful or illicit searches does not amount to "development" for purposes of the immunity exception. A dating website that requires users to enter their sex, race, religion and marital status through drop-down menus, and that provides means for users to search along the same lines, retains its CDA immunity insofar as it does not contribute to any alleged illegality;^[23] this immunity is retained even if the website is sued for libel based on these characteristics because the website would not have contributed materially to any alleged defamation. Similarly, a housing website that allows users to specify whether they will or will not receive emails by means of user-defined criteria might help some users exclude email from other users of a particular race or sex. However, that website would be immune, so long as it does not require the use of discriminatory criteria. A website operator who edits user-created content — such as by correcting spelling, removing obscenity or trimming for length — retains his "immunity for any illegality in the user-created content, provided that the edits are unrelated to the illegality. However, a website operator who edits in a manner that contributes to the alleged illegality — such as by removing the word "not" from a user's message reading "[Name] did not steal the artwork" in order to transform an innocent message into a libelous one — is directly involved in the alleged illegality and thus not immune.^[24]

Here, Roommate's connection to the discriminatory filtering process is direct and palpable: Roommate designed its search and email systems to limit the listings available to subscribers based on sex, sexual orientation and presence of children.^[25] Roommate selected the criteria used to hide listings, and Councils allege that the act of hiding certain listings is itself unlawful under the Fair Housing Act, which prohibits brokers from steering clients in accordance with discriminatory [1170] preferences.^[26] We need not decide the merits of Councils' claim to hold that Roommate is sufficiently involved with the design and operation of the search and email systems — which are engineered to limit access to housing on the basis of the protected characteristics elicited by the registration process — so as to forfeit any immunity to which it was otherwise entitled under section 230.

Roommate's situation stands in stark contrast to Stratton Oakmont, the case Congress sought to reverse through passage of section 230. There, defendant Prodigy was held liable for a user's unsolicited message because it attempted to remove some problematic content from its website, but didn't remove enough. Here, Roommate is not being sued for removing some harmful messages while failing to remove others; instead, it is being sued for the predictable consequences of creating a website designed to solicit and enforce housing preferences that are alleged to be illegal.

We take this opportunity to clarify two of our previous rulings regarding the scope of section 230 immunity. Today's holding sheds additional light on Batzel v. Smith, 333 F.3d 1018 (9th Cir.2003). There, the editor of an email newsletter received a tip about some artwork, which the tipster falsely alleged to be stolen. The newsletter editor incorporated the tipster's email into the next issue of his newsletter and added a short headnote, which he then emailed to his subscribers.^[27] The art owner sued for libel and a split panel held the newsletter editor to be immune under section 230 of the CDA.^[28]

Our opinion is entirely consistent with that part of Batzel which holds that an editor's minor changes to the spelling, grammar and length of third-party content do not strip him of section 230 immunity. None of those changes contributed to the libelousness of the message, so they do not add up to "development" as we interpret the term. See pp. 1167-69 supra. Batzel went on to hold that the editor could be liable for selecting the tipster's email for inclusion in the newsletter, depending on whether or not the tipster had tendered the piece to the editor for posting online, and remanded for a determination of that issue. Batzel, 333 F.3d at 1035.

The distinction drawn by Batzel anticipated the approach we take today. As Batzel explained, if the tipster tendered the material for posting online, then the editor's job was, essentially, to determine whether or not to prevent its posting — precisely the kind of activity for which section 230 was meant to provide immunity.^[29] And any activity that can be boiled [1171] down to deciding whether to exclude material that third parties seek to post online is perforce immune under section 230. See p. 1171-72 & n. 32 infra. But if the editor publishes material that he does not believe was tendered to him for posting online, then he is the one making the affirmative decision to publish, and so he contributes materially to its allegedly unlawful dissemination. He is thus properly deemed a developer and not entitled to CDA immunity. See Batzel, 333 F.3d at 1033.^[30]

We must also clarify the reasoning undergirding our holding in Carafano v. Metrosplash.com, Inc., 339 F.3d 1119 (9th Cir.2003), as we used language there that was unduly broad. In Carafano, an unknown prankster impersonating actress Christianne Carafano created a profile for her on an online dating site. The profile included Carafano's home address and suggested that she was looking for an unconventional liaison. When Carafano received threatening phone calls, she sued the dating site for publishing the unauthorized profile. The site asserted immunity under section 230. We correctly held that the website was immune, but incorrectly suggested that it could never be liable because "no [dating] profile has any content until a user actively creates it." Id. at 1124. As we explain above, see pp. 1165-69 supra, even if the data are supplied by third parties, a website operator may still contribute to the content's illegality and thus be liable as a developer.^[31] Providing immunity every time a website uses data initially obtained from third parties would eviscerate the exception to section 230 for "develop[ing]" unlawful content "in whole or in part." 47 U.S.C. § 230(f)(3).

We believe a more plausible rationale for the unquestionably correct result in Carafano is this: The allegedly libelous content there — the false implication that Carafano was unchaste — was created and developed entirely by the malevolent user, without prompting or help from the website operator. To be sure, the website provided neutral tools, which the anonymous dastard used to publish the libel, but the website did absolutely nothing to encourage the posting of defamatory content — indeed, the defamatory posting was contrary to the website's express policies. The claim against the website was, in effect, that it failed to review each user-created profile to ensure that it wasn't defamatory. That is precisely the kind of [1172] activity for which Congress intended to grant absolution with the passage of section 230. With respect to the defamatory content, the website operator was merely a passive conduit and thus could not be held liable for failing to detect and remove it.^[32]

By contrast, Roommate both elicits the allegedly illegal content and makes aggressive use of it in conducting its business. Roommate does not merely provide a framework that could be utilized for proper or improper purposes; rather, Roommate's work in developing the discriminatory questions, discriminatory answers and discriminatory search mechanism is directly related to the alleged illegality of the site. Unlike Carafano, where the website operator had nothing to do with the user's decision to enter a celebrity's name and personal information in an otherwise licit dating service, here, Roommate is directly involved with developing and enforcing a system that subjects subscribers to allegedly discriminatory housing practices.

Our ruling today also dovetails with another facet of Carafano: The mere fact that an interactive computer service "classifies user characteristics . . . does not transform [it] into a `developer' of the `underlying misinformation.'" Carafano, 339 F.3d at 1124. Carafano, like Batzel, correctly anticipated our common-sense interpretation of the term "develop[]" in section 230. Of course, any classification of information, like the sorting of dating profiles by the type of relationship sought in Carafano, could be construed as "develop[ment]" under an unduly broad reading of the term. But, once again, such a broad reading would sap section 230 of all meaning.

The salient fact in Carafano was that the website's classifications of user characteristics did absolutely nothing to enhance the defamatory sting of the message, to encourage defamation or to make defamation easier: The site provided neutral tools specifically designed to match romantic partners depending on their voluntary inputs. By sharp contrast, Roommate's website is designed to force subscribers to divulge protected characteristics and discriminatory preferences, and to match those who have rooms with those who are looking for rooms based on criteria that appear to be prohibited by the FHA.^[33]

[1173] 3. Councils finally argue that Roommate should be held liable for the discriminatory statements displayed in the "Additional Comments" section of profile pages. At the end of the registration process, on a separate page from the other registration steps, Roommate prompts subscribers to "tak[e] a moment to personalize your profile by writing a paragraph or two describing yourself and what you are looking for in a roommate." The subscriber is presented with a blank text box, in which he can type as much or as little about himself as he wishes. Such essays are visible only to paying subscribers.

Subscribers provide a variety of provocative, and often very revealing, answers. The contents range from subscribers who "[p]ref[er] white Male roommates" or require that "[t]he person applying for the room MUST be a BLACK GAY MALE" to those who are "NOT looking for black muslims." Some common themes are a desire to live without "drugs, kids or animals" or "smokers, kids or druggies," while a few subscribers express more particular preferences, such as preferring to live in a home free of "psychos or anyone on mental medication." Some subscribers are just looking for someone who will get along with their significant other^[34] or with their most significant Other.^[35]

Roommate publishes these comments as written.^[36] It does not provide any specific guidance as to what the essay should contain, nor does it urge subscribers to input [1174] discriminatory preferences. Roommate is not responsible, in whole or in part, for the development of this content, which comes entirely from subscribers and is passively displayed by Roommate. Without reviewing every essay, Roommate would have no way to distinguish unlawful discriminatory preferences from perfectly legitimate statements. Nor can there be any doubt that this information was tendered to Roommate for publication online. See pp. 1170-71 supra. This is precisely the kind of situation for which section 230 was designed to provide immunity. See pp. 1162-64 supra.

The fact that Roommate encourages subscribers to provide something in response to the prompt is not enough to make it a "develop[er]" of the information under the common-sense interpretation of the term we adopt today. It is entirely consistent with Roommate's business model to have subscribers disclose as much about themselves and their preferences as they are willing to provide. But Roommate does not tell subscribers what kind of information they should or must include as "Additional Comments," and certainly does not encourage or enhance any discriminatory content created by users. Its simple, generic prompt does not make it a developer of the information posted.^[37]

Councils argue that — given the context of the discriminatory questions presented earlier in the registration process — the "Additional Comments" prompt impliedly suggests that subscribers should make statements expressing a desire to discriminate on the basis of protected classifications; in other words, Councils allege that, by encouraging some discriminatory preferences, Roommate encourages other discriminatory preferences when it gives subscribers a chance to describe themselves. But the encouragement that bleeds over from one part of the registration process to another is extremely weak, if it exists at all. Such weak encouragement cannot strip a website of its section 230 immunity, lest that immunity be rendered meaningless as a practical matter.^[38]

We must keep firmly in mind that this is an immunity statute we are expounding, a provision enacted to protect websites against the evil of liability for failure to remove offensive content. See pp. 1162-64 supra. Websites are complicated enterprises, and there will always be close cases where a clever lawyer could argue that something the website operator did encouraged the illegality. Such close cases, we believe, must be resolved in favor of immunity, lest we cut the heart out of section 230 by forcing websites to face death by ten thousand duck-bites, fighting off claims that they promoted or encouraged — or at least tacitly assented to — the illegality of third parties. Where it is very clear that the website directly participates in developing the alleged illegality — as it is clear here with respect to Roommate's questions, answers and the resulting profile pages — immunity will be lost. But in cases of enhancement by implication or [1175] development by inference — such as with respect to the "Additional Comments" here — section 230 must be interpreted to protect websites not merely from ultimate liability, but from having to fight costly and protracted legal battles.

The dissent prophesies doom and gloom for countless Internet services, Dissent at 1183-84, but fails to recognize that we hold part of Roommate's service entirely immune from liability. The search engines the dissent worries about, id., closely resemble the "Additional Comments" section of Roommate's website. Both involve a generic text prompt with no direct encouragement to perform illegal searches or to publish illegal content. We hold Roommate immune and there is no reason to believe that future courts will have any difficulty applying this principle.^[39] The message to website operators is clear: If you don't encourage illegal content, or design your website to require users to input illegal content, you will be immune.

We believe that this distinction is consistent with the intent of Congress to preserve the free-flowing nature of Internet speech and commerce without unduly prejudicing the enforcement of other important state and federal laws. When Congress passed section 230 it didn't' intend to prevent the enforcement of all laws online; rather, it sought to encourage interactive computer services that provide users neutral tools to post content online to police that content without fear that through their "good Samaritan . . . screening of offensive material," 47 U.S.C. § 230(c), they would become liable for every single message posted by third parties on their website.

* * *

In light of our determination that the CDA does not provide immunity to Roommate for all of the content of its website and email newsletters, we remand for the district court to determine in the first instance whether the alleged actions for which Roommate is not immune violate the Fair Housing Act, 42 U.S.C. § 3604(c).^[40] We vacate the dismissal of the state law claims so that the district court may reconsider whether to exercise its supplemental jurisdiction in light of our ruling on the federal claims. Fredenburg v. Contra Costa County Dep't of Health Servs., 172 F.3d 1176, 1183 (9th Cir.1999). We deny Roommate's [1176] cross-appeal of the denial of attorneys' fees and costs; Councils prevail on some of their arguments before us so their case is perforce not frivolous.

REVERSED in part, VACATED in part, AFFIRMED in part and REMANDED. NO COSTS.

McKEOWN, Circuit Judge, with whom RYMER and BEA, Circuit Judges, join, concurring in part and dissenting in part:

The ubiquity of the Internet is undisputed. With more than 1.3 billion Internet users and over 158 million websites in existence,^[41] a vast number of them interactive like Google, Yahoo!, Craigslist, MySpace, YouTube, and Facebook, the question of webhost liability is a significant one. On a daily basis, we rely on the tools of cyberspace to help us make, maintain, and rekindle friendships; find places to live, work, eat, and travel; exchange views on topics ranging from terrorism to patriotism; and enlighten ourselves on subjects from "aardvarks to Zoroastrianism."^[42]

The majority's unprecedented expansion of liability for Internet service providers threatens to chill the robust development of the Internet that Congress envisioned. The majority condemns Roommate's "search system," a function that is the heart of interactive service providers. My concern is not an empty Chicken Little "sky is falling" alert. By exposing every interactive service provider to liability for sorting, searching, and utilizing the all too familiar drop-down menus, the majority has dramatically altered the landscape of Internet liability. Instead of the "robust"^[43] immunity envisioned by Congress, interactive service providers are left scratching their heads and wondering where immunity ends and liability begins.

To promote the unfettered development of the Internet, Congress adopted the Communications Decency Act of 1996 ("CDA"), which provides that interactive computer service providers will not be held legally responsible for publishing information provided by third parties. 47 U.S.C. § 230(c)(1). Even though traditional publishers retain liability for performing essentially equivalent acts in the "non-virtual world," Congress chose to treat interactive service providers differently by immunizing them from liability stemming from sorting, searching, and publishing third-party information. As we explained in Batzel v. Smith:

[Section] 230(c)(1)[] overrides the traditional treatment of publishers, distributors, and speakers under statutory and common law. As a matter of policy, "Congress decided not to treat providers of interactive computer services like other information providers such as newspapers, magazines or television and radio stations. . . ." Congress . . . has chosen to treat cyberspace differently.

333 F.3d 1018, 1026-1027 (9th Cir.2003) (quoting Blumenthal v. Drudge, 992 F.Supp. 44, 49 (D.D.C.1998) (footnote omitted)).

Now, with the stroke of a pen or, more accurately, a few strokes of the keyboard, the majority upends the settled view that interactive service providers enjoy broad immunity when publishing information provided by third parties. Instead, interactive [1177] service providers are now joined at the hip with third-party users, and they rise and fall together in liability for Internet sortings and postings.

To be sure, the statute, which was adopted just as the Internet was beginning a surge of popular currency,^[44] is not a perfect match against today's technology. The Web 2.0 version is a far cry from web technology in the mid-1990s. Nonetheless, the basic message from Congress has retained its traction, and there should be a high bar to liability for organizing and searching third-party information. The bipartisan view in Congress was that the Internet, as a new form of communication, should not be impeded by the transference of regulations and principles developed from traditional modes of communication. The majority repeatedly harps that if something is prohibited in the physical world, Congress could not have intended it to be legal in cyberspace. Yet that is precisely the path Congress took with the CDA: the anomaly that a webhost may be immunized for conducting activities in cyberspace that would traditionally be cause for liability is exactly what Congress intended by enacting the CDA.

In the end, the majority offers interactive computer service providers no bright lines and little comfort in finding a home within § 230(c)(1). The result in this case is driven by the distaste for housing discrimination, a laudable endgame were housing the real focus of this appeal. But it is not. I share the majority's view that housing discrimination is a troubling issue. Nevertheless, we should be looking at the housing issue through the lens of the Internet, not from the perspective of traditional publisher liability. Whether § 230(c)(1) trumps the Fair Housing Act ("FHA") is a policy decision for Congress, not us. Congress has spoken: third-party content on the Internet should not be burdened with the traditional legal framework.

I respectfully part company with the majority as to Part 2^[45] of the opinion because the majority has misconstrued the statutory protection under the CDA for Roommate's publishing and sorting of user profiles. The plain language and structure of the CDA unambiguously demonstrate that Congress intended these activities — the collection, organizing, analyzing, searching, and transmitting of third-party content — to be beyond the scope of traditional publisher liability. The majority's decision, which sets us apart from five circuits, contravenes congressional intent and violates the spirit and serendipity of the Internet.

Specifically, the majority's analysis is flawed for three reasons: (1) the opinion conflates the questions of liability under the FHA and immunity under the CDA; (2) the majority rewrites the statute with its definition of "information content provider," labels the search function "information development," and strips interactive service providers of immunity; and (3) the majority's approach undermines the purpose [1178] of § 230(c)(1) and has far-reaching practical consequences in the Internet world.

To begin, it is important to recognize what this appeal is not about. At this stage, there has been no determination of liability under the FHA, nor has there been any determination that the questions, answers or even the existence of Roommate's website violate the FHA. The FHA is a complicated statute and there may well be room for potential roommates to select who they want to live with, e.g., a tidy accountant wanting a tidy professional roommate, a collegiate male requesting a male roommate, an observant Jew needing a house with a kosher kitchen, or a devout, single, religious female preferring not to have a male housemate. It also bears noting that even if Roommate is immune under the CDA, the issue of user liability for allegedly discriminatory preferences is a separate question. See Zeran v. Am. Online, Inc., 129 F.3d 327, 330 (4th Cir. 1997) (stating that "the"original culpable party" does not "escape accountability").

By offering up inflammatory examples, the majority's opinion screams "discrimination." The hazard is, of course, that the question of discrimination has not yet been litigated. In dissenting, I do not condone housing discrimination or endorse unlawful discriminatory roommate selection practices; I simply underscore that the merits of the FHA claim are not before us. However, one would not divine this posture from the majority's opinion, which is infused with condemnation of Roommate's users' practices. To mix and match, as does the majority, the alleged unlawfulness of the information with the question of webhost immunity is to rewrite the statute.

Examples from the opinion highlight that the majority's conclusion rests on the premise that Roommate's questions and matching function violate the FHA:

• "Unlawful questions solicit (a.k.a. `develop') unlawful answers." Maj. Op. at 1166.

• "If such questions are unlawful when posed face-to-face or by telephone, they don't magically become lawful when asked electronically online." Id. at 1164.

• "If such screening is prohibited when practiced in person or by telephone, we see no reason why Congress would have wanted to make it lawful to profit from it online." Id. at 1167.

• "Roommate's search function thus differs materially from generic search engines such as Google, Yahoo! and MSN Live Search, in that Roommate designed its system to use allegedly unlawful criteria so as to limit the results of each search, and to force users to participate in its discriminatory process." Id.

• "By contrast, ordinary search engines do not use unlawful criteria to limit the scope of searches conducted on them, nor are they designed to achieve illegal ends — as Roommate's search function is alleged to do here." Id.

• "Roommate's website is designed to force subscribers to divulge protected characteristics and discriminatory preferences." Id. at 1172.

The entire opinion links Roommate's ostensibly reprehensible conduct (and that of its users) with an unprecedented interpretation of the CDA's immunity provision. The majority condemns Roommate for soliciting illegal content, but there has been no determination that Roommate's questions or standardized answers are illegal. Instead of foreshadowing a ruling on the FHA, the opinion should be confined to the issue before us — application of § 230(c)(1) to Roommate. The district court has not yet ruled on the merits of the FHA claim and neither should we.

[1179] The Statute

With this background in mind, I first turn to the text of the statute. Section 230 begins with a detailed recitation of findings and policy reasons for the statute. Congress expressly found that the "Internet and other interactive computer services offer a forum for a true diversity of political discourse, unique opportunities for cultural development, and myriad avenues for intellectual activity," and that "[i]ncreasingly Americans are relying on interactive media for a variety of political, educational, cultural, and entertainment services." 47 U.S.C. § 230(a)(3), (5). Congress declared that "[i]t is the policy of the United States to . . . promote the continued development of the Internet and other interactive computer services and other interactive media." § 230(b)(1).^[46]

Unlike some statutes, subsections (a) and (b) set out in clear terms the congressional findings and policies underlying the statute. For this reason, it strikes me as odd that the majority begins, not with the statute and these express findings, but with legislative history. Granted, Congress was prompted by several cases, particularly the Prodigy case, to take action to protect interactive service providers. See Stratton Oakmont, Inc. v. Prodigy Servs. Co., 1995 WL 323710, 1995 N.Y. Misc. LEXIS 229 (N.Y.Sup.Ct. May 24, 1995). But that case does not cabin the scope of the statute, and the background leading up to enactment of the CDA is no substitute for the language of the statute itself. See Chicago Lawyers' Comm. for Civil Rights Under the Law, Inc. v. Craigslist, Inc., 519 F.3d 666, ____ (7th Cir. 2008) (concluding that, as enacted, "Section 230(c)(1) is general[,]" despite its "genesis" in Prodigy).

Section 230(c), the heart of this case, is entitled "Protection for `good Samaritan' blocking and screening of offensive material[.]" The substantive language of the statute itself is not so limited. Section 230(c)(1) provides:

(1) Treatment of publisher or speaker

No provider or user of an interactive computer service shall be treated as the publisher or speaker of any information provided by another information content provider.

§ 230(c)(1). Since it was first addressed in 1997 in Zeran, this section has been interpreted by the courts as providing webhost "immunity," although to be more precise, it provides a safe haven for interactive computer service providers by removing them from the traditional liabilities attached to speakers and publishers.^[47]See Zeran, 129 F.3d at 330 ("By its plain language, § 230 creates a federal immunity to any cause of action that would make service providers liable for information originating with a third-party user of the service.").

We have characterized this immunity under § 230(c)(1) as "quite robust." Carafano, 339 F.3d at 1123. Five of our sister circuits have similarly embraced this robust view of immunity by providing differential treatment to interactive service providers. Chicago Lawyers' Comm. for Civil Rights Under the Law, Inc. v. Craigslist, Inc., 519 F.3d 666, ___ _ ___ (7th Cir. 2008); Universal Commc'n Sys. v. Lycos, [1180] Inc., 478 F.3d 413, 415 (1st Cir.2007); Green v. Am. Online, 318 F.3d 465, 470 (3d Cir.2003); Ben Ezra, Weinstein, & Co., Inc. v. Am. Online Inc., 206 F.3d 980, 986 (10th Cir.2000); Zeran, 129 F.3d at 330; see also Whitney Info. Network, Inc. v. Xcentric Ventures, LLC, No. 2:04-cv-47-FtM-34SPC, 2008 WL 450095, 2008 U.S. Dist. LEXIS 11632 (M.D.Fla. Feb. 15, 2008); Doe v. MySpace, Inc., 474 F.Supp.2d 843, 849 (W.D.Tex.2007); Corbis Corp. v. Amazon.com, Inc., 351 F.Supp.2d 1090, 1118 (W.D.Wash.2004); Blumenthal, 992 F.Supp. at 50-53; Barrett v. Rosenthal, 40 Cal.4th 33, 51 Cal. Rptr.3d 55, 146 P.3d 510, 529 (2006); Gentry v. eBay, Inc., 99 Cal.App.4th 816, 121 Cal.Rptr.2d 703, 717-18 (2002); Schneider v. Amazon.com, Inc., 108 Wash.App. 454, 31 P.3d 37, 42-43 (2001).

Key to this immunity provision are the terms "interactive computer service" provider and "information content provider." The CDA defines an "interactive computer service" as any "information service, system, or access software provider that provides or enables computer access by multiple users to a computer server." § 230(f)(2). An interactive computer service provider is not liable as a "publisher" or "speaker" of information if the "information" is "provided by another information content provider." § 230(c)(1). The statute then defines an "information content provider" as a "person or entity that is responsible, in whole or in part, for the creation or development of information provided through the Internet or any other interactive computer service." § 230(f)(3). If the provider of an interactive computer service is an information content provider of the information at issue, it cannot claim immunity as a publisher or speaker. Carafano, 339 F.3d at 1123.

Courts deciding the question of § 230(c)(1) immunity "do not write on a blank slate." Universal Commc'n, 478 F.3d at 418. Even though rapid developments in technology have made webhosts increasingly adept at searching and displaying third-party information, reviewing' courts have, in the twelve years since the CDA's enactment, "adopt[ed] a relatively expansive definition of `interactive computer service' and a relatively restrictive definition of `information content provider.'" See Carafano, 339 F.3d at 1123 (footnotes omitted). As long as information is provided by a third party, webhosts are immune from liability for publishing "ads for housing, auctions of paintings that may have been stolen by Nazis, biting comments about steroids in baseball, efforts to verify the truth of politicians' promises, and everything else that third parties may post on a web site." Craigslist, 519 F.3d at ____. We have underscored that this broad grant of webhost immunity gives effect to Congress's stated goals "to promote the continued development of the Internet and other interactive computer services" and "to preserve the vibrant and competitive free market that presently exists for the Internet and other interactive computer services." Carafano, 339 F.3d at 1123 (discussing § 230(b)(1), (2)).

Application of § 230(c)(1) to Roommate's Website

Because our focus is on the term "information content provider," and what it means to create or develop information, it is worth detailing exactly how the website operates, what information is at issue and who provides it. The roommate matching process involves three categories of data: About Me or Household Description; Roommate Preferences; and Comments.

To become a member of Roommates.com, a user must complete a personal profile by selecting answers from drop-down menus or checking off boxes on the screen. The profile includes "location" information [1181] (e.g., city and state, region of the city, and data about the surrounding neighborhood); details about the residence (e.g., the total number of bedrooms and bathrooms in the home, and amenities such as air conditioning, wheelchair access, high-speed Internet, or parking), and the "rental details" (e.g., monthly rent charged, lease period, and availability). The last section of the profile is the "Household Description" section,^[48] which includes the total number of occupants in the home, their age range, gender, occupation, level of cleanliness, whether they are smokers, and whether children or pets are present.

The remaining sections of the registration process are completely optional; a user who skips them has created a profile based on the information already provided. At his option, the user may select an emoticon to describe the "household character," and may upload images of the room or residence. Next, users may, at their option, specify characteristics desired in a potential roommate, such as a preferred age range, gender, and level of cleanliness. If nothing is selected, all options are included.^[49] The final step in the registration process, which is also optional, is the "Comments" section, in which users are presented with a blank text box in which they may write whatever they like, to be published with their member profiles.

Users may choose an optional "custom search" of user profiles based on criteria that they specify, like the amount of monthly rent or distance from a preferred city. Based on the information provided by users during the registration process, Roommate's automated system then searches and matches potential roommates. Roommate's Terms of Service provide in part, "You understand that we do not provide the information on the site and that all publicly posted or privately transmitted information, data, text, photographs, graphics, messages, or other materials (`Content') are the sole responsibility of the person from which such Content originated."

Roommate's users are "information content providers" because they are responsible for creating the information in their user profiles and, at their option — not the website's choice — in expressing preferences as to roommate characteristics. § 230(f)(3). The critical question is whether Roommate is itself an "information content provider," such that it cannot claim that the information at issue was "provided [1182] by another information content provider." A close reading of the statute leads to the conclusion that Roommate is not an information content provider for two reasons: (1) providing a drop-down menu does not constitute "creating" or "developing" information; and (2) the structure and text of the statute make plain that Congress intended to immunize Roommate's sorting, displaying, and transmitting of third-party information.

Roommate neither "creates" nor "develops" the information that is challenged by the Councils, i.e., the information provided by the users as to their protected characteristics and the preferences expressed as to roommate characteristics. All Roommate does is to provide a form with options for standardized answers. Listing categories such as geographic location, cleanliness, gender and number of occupants, and transmitting to users profiles of other users whose expressed information matches their expressed preferences, can hardly be said to be creating or developing information. Even adding standardized options does not "develop" information. Roommate, with its prompts, is merely "selecting material for publication," which we have stated does not constitute the "development" of information. Batzel, 333 F.3d at 1031. The profile is created solely by the user, not the provider of the interactive website. Indeed, without user participation, there is no information at all. The drop-down menu is simply a precategorization of user information before the electronic sorting and displaying that takes place via an algorithm. If a user has identified herself as a non-smoker and another has expressed a preference for a non-smoking roommate, Roommate's sorting and matching of user information are no different than that performed by a generic search engine.

Displaying the prompt "Gender" and offering the list of choices, "Straight male; Gay male; Straight female; Gay female" does not develop the information, "I am a Gay male." The user has identified himself as such and provided that information to Roommate to publish. Thus, the user is the sole creator of that information; no "development" has occurred. In the same vein, presenting the user with a "Preferences" section and drop-down menus of options does not "develop" a user's preference for a non-smoking roommate. As we stated in Carafano, the "actual profile `information' consist[s] of the particular options chosen" by the user, such that Roommate is not "responsible, even in part, for associating certain multiple choice responses with a set of [] characteristics." 339 F.3d at 1124.

The thrust of the majority's proclamation that Roommate is "developing" the information that it publishes, sorts, and transmits is as follows: "[W]e interpret the term `development' as referring not merely to augmenting the content generally, but to materially contributing to its unlawfulness." Maj. Op. at 1168. This definition is original to say the least and springs forth untethered to anything in the statute.

The majority's definition of "development" epitomizes its consistent collapse of substantive liability with the issue of immunity. Where in the statute does Congress say anything about unlawfulness? Whether Roommate is entitled to immunity for publishing and sorting profiles is wholly distinct from whether Roommate may be liable for violations of the FHA. Immunity has meaning only when there is something to be immune from, whether a disease or the violation of a law. It would be nonsense to claim to be immune only from the innocuous. But the majority's immunity analysis is built on substantive liability: to the majority, CDA immunity depends on whether a webhost materially [1183] contributed to the unlawfulness of the information. Whether the information at issue is unlawful and whether the webhost has contributed to its unlawfulness are issues analytically independent of the determination of immunity. Grasping at straws to distinguish Roommate from other interactive websites such as Google and Yahoo!, the majority repeatedly gestures to Roommate's potential substantive liability as sufficient reason to disturb its immunity. But our task is to determine whether the question of substantive liability may be reached in the first place.

Keep in mind that "unlawfulness" would include not only purported statutory violations but also potential defamatory statements. The irony is that the majority would have us determine "guilt" or liability in order to decide whether immunity is available. This upside-down approach would knock out even the narrowest immunity offered under § 230(c) — immunity for defamation as a publisher or speaker.

Another flaw in the majority's approach is that it fails to account for all of the other information allegedly developed by the webhost. For purposes of determining whether Roommate is an information content provider vis-a-vis the profiles, the inquiry about geography and the inquiry about gender should stand on the same footing. Both are single word prompts followed by a drop-down menu of options. If a prompt about gender constitutes development, then so too does, the prompt about geography. And therein lies the rub.

Millions of websites use prompts and drop-down menus. Inquiries range from what credit card you want to use and consumer satisfaction surveys asking about age; sex and household income, to dating sites, e.g., match.com, sites lambasting corporate practices, e.g., ripoffreports.com, and sites that allow truckers to link up with available loads, e.g., getloaded.com. Some of these sites are innocuous while others may not be. Some may solicit illegal information; others may not. But that is not the point. The majority's definition of "development" would transform every interactive site into an information content provider and the result would render illusory any immunity under § 230(c). Virtually every site could be responsible in part for developing content.

For example, the majority purports to carve out a place for Google and other search engines. Maj. Op. at 1167. But the modern Google is more than a match engine: it ranks search results, provides prompts beyond what the user enters, and answers questions. In contrast, Roommate is a straight match service that searches information and criteria provided by the user, not Roommate. It should be afforded no less protection than Google, Yahoo!, or other search engines.

The majority then argues that "providing neutral tools to carry out what may be unlawful or illicit searches does not amount to `development.'" Maj. Op. at 1169. But this effort to distinguish Google, Yahoo!, and other search engines from Roommate is unavailing. Under the majority's definition of "development," these search engines are equivalent to Roommate. Google "encourages" or "contributes" (the majority's catch phrases) to the unlawfulness by offering search tools that allow the user to perform an allegedly unlawful match. If a user types into Google's search box, "looking for a single, Christian, female roommate," and Google displays responsive listings, Google is surely "materially contributing to the alleged unlawfulness" of information created by third parties, by publishing their intention to discriminate on the basis of protected characteristics. In the defamation arena, a webhost's publication of a defamatory statement "materially contributes" to its [1184] unlawfulness, as publication to third parties is an element of the offense. At bottom, the majority's definition of "development" can be tucked in, let out, or hemmed up to fit almost any search engine, creating tremendous uncertainty in an area where Congress expected predictability.

"Development" is not without meaning. In Batzel, we hinted that the "development of information" that transforms one into an "information content provider" is "something more substantial than merely editing portions of an email and selecting material for publication." 333 F.3d at 1031. We did not flesh out further the meaning of "development" because the editor's alterations of an email message and decision to publish it did not constitute "development." Id.

Because the statute does not define "development," we should give the term its ordinary meaning. See San Jose Christian Coll. v. City of Morgan Hill, 360 F.3d 1024, 1034 (9th Cir.2004) (stating that dictionaries may be used to determine the "`plain meaning' of a term undefined by a statute"). "Development" is defined in Webster's Dictionary as a "gradual advance or growth through progressive changes." Webster's Third New International Dictionary 618 (2002). The multiple uses of "development" and "develop" in other provisions of § 230 give texture to the definition of "development," and further expose the folly of the majority's ungrounded definition. See, e.g., § 230(b)(3) (stating that "[i]t is the policy of the United States to encourage the development of technologies which maximize user control over what information is received by individuals, families, and' schools") (emphasis added).^[50] Defining "development" in this way keeps intact the settled rule that the CDA immunizes a webhost who exercises a publisher's "traditional editorial functions — such as deciding whether to publish, withdraw, post-pone, or alter content." Batzel, 333 F.3d at 1031 n. 18.^[51]

Applying the plain meaning of "development" to Roommate's sorting and transmitting of third-party information demonstrates [1185] that it was not transformed into an "information content provider." In searching, sorting, and transmitting information, Roommate made no changes to the information provided to it by users. Even having notice that users may be using its site to make discriminatory statements is not sufficient to invade Roommate's immunity. See Zeran, 129 F.3d at 333 (stating that "liability upon notice has a chilling effect on the freedom of Internet speech.").

The majority blusters that Roommate develops information, because it "requir[es] subscribers to provide the information as a condition of accessing its services," and "designed its search system so it would steer users based on the preferences and personal characteristics that Roommate itself forces subscribers to disclose." Maj. Op. at 1165, 1167.^[52] But the majority, without looking back, races past the plain language of the statute. That Roommate requires users to answer a set of prompts to identify characteristics about themselves does not change the fact that the users have furnished this information to Roommate for Roommate to publish in their profiles. Nor do Roommate's prompts alter the fact that users have chosen to select characteristics that they find desirable in potential roommates, and have directed Roommate to search and compile results responsive to their requests. Moreover, tagging Roommate with liability for the design of its search system is dangerous precedent for analyzing future Internet cases.

Even if Roommate's prompts and drop-down menus could be construed to seek out, or encourage, information from users, the CDA does not withhold immunity for the encouragement or solicitation of information.^[53]See Blumenthal, 992 F.Supp. at 52 (stating that "Congress has made a different policy choice by providing immunity even where the interactive service provider has an active, even aggressive role in making available content prepared by others.") (emphasis added); Gentry, 121 Cal.Rptr.2d at 718 (noting that "enforcing appellants' negligence claim would place liability on eBay for simply compiling false and/or misleading content created by the individual defendants and other coconspirators."). The CDA does not countenance an exception for the solicitation or encouragement of information provided by users.

A number of district courts have recently encountered the claim that an interactive website's solicitation of information, by requiring user selection of content from drop-down menus, transformed it into an information content provider. Unsurprisingly, these courts reached the same commonsense solution that I reach here: § 230(c)(1) immunizes the interactive service provider. See Whitney Info. Network, Inc. v. Xcentric Ventures, LLC, No. 2:04-cv-47-FtM-34SPC, 2008 WL 450095, at *10, 2008 U.S. Dist. LEXIS 11632, at *36 (M.D.Fla. Feb. 15, 2008) (stating that the "mere fact that Xcentric provides categories from which a poster must make a selection in order to submit a report on the [] website is not sufficient to treat Defendants as information content providers of the reports"); Global Royalties, Ltd. v. Xcentric Ventures, LLC, No. 07-956-PHX-FJM, [1186] 2007 WL 2949002, 2007 U.S. Dist. LEXIS 77551 (D.Ariz. Oct. 10, 2007). Simply supplying a list of options from which a user must select options "is minor and passive participation" that does not defeat CDA immunity. Global Royalties, 2007 WL 2949002, at *3, 2007 U.S. Dist. LEXIS 77551, at *9; see also Corbis, 351 F.Supp.2d at 1118 (holding that even though Amazon.com "may have encouraged third parties to use the Zshops platform and provided the tools to assist them, that does not disqualify it from immunity under § 230 because the Zshops vendor ultimately decided what information to put on its site.").

Carafano presented circumstances virtually indistinguishable from those before us, yet the majority comes to the exact opposite conclusion here in denying immunity for sorting and matching third-party information provided in response to webhost prompts. The website in Carafano, an online dating service named Matchmaker.com, asked its users sixty-two detailed questions and matched users according to their responses. We held that § 230(c)(1) immunized the dating service, and flatly rejected the proposition that matching, sorting, and publishing user information in response to webhost prompts abrogated CDA immunity. Carafano, 339 F.3d at 1124-25. A provider's "decision to structure the information provided by users," which enables the provider to "offer additional features, such as `matching' profiles with similar characteristics or highly structured searches based on combinations of multiple choice questions," ultimately "promotes the expressed Congressional policy `to promote the continued development of the Internet and other interactive computer services.'" Id. (quoting § 230(b)(1)). Now the majority narrows Carafano on the basis that Matchmaker did not prompt the allegedly libelous information that was provided by a third party. Maj. Op. at 1171. But the majority makes this distinction without any language in the statute supporting the consideration of the webhost's prompting or solicitation.

The structure of the statute also supports my view that Congress intended to immunize Roommate's sorting and publishing of user profiles. An "interactive computer service" is defined to include an "access software provider." § 230(f)(2). The statute defines an "access software provider" as one that provides "enabling tools" to "filter," "screen," "pick," "choose," "analyze," "digest," "search," "forward," "organize," and "reorganize" content. § 230(f)(4)(A)-(C).

By providing a definition for "access software provider" that is distinct from the definition of an "information content provider," and withholding immunity for "information content providers," the statute makes resoundingly clear that packaging, sorting, or publishing third-party information are not the kind of activities that Congress associated with "information content providers." Yet these activities describe exactly what Roommate does through the publication and distribution of user profiles: Roommate "receives," "filters," "digests," and "analyzes" the information provided by users in response to its registration prompts, and then "transmits," "organizes," and "forwards" that information to users in the form of uniformly organized profiles. Roommate is performing tasks that Congress recognized as typical of entities that it intended to immunize.

Finally, consider the logical disconnect of the majority's opinion. The majority writes — and I agree — that the openended Comments section contains only third-party content. Maj. Op. at 1173-75. But if Roommate's search function permits sorting by key words such as children or gender, the majority would label Roommate's use of such criteria as a "discriminatory filtering process." Id. at 1169-70.

[1187] At a minimum, the CDA protects the search criteria employed by websites and does not equate tools that "filter," "screen," "pick," "choose," "analyze," "digest," "search," "forward," "organize," and "reorganize" with the "creation or development" of information. § 230(f)(4)(A)-(C).

Ramifications of the Majority Opinion

I am troubled by the consequences that the majority's conclusion poses for the ever-expanding Internet community. The unwise narrowing of our precedent, coupled with the mixing and matching of CDA immunity with substantive liability, make it exceedingly difficult for website providers to know whether their activities will be considered immune under the CDA. We got it right in Carafano, that "[u]nder § 230(c) . . . so long as a third party willingly provides the essential published content, the interactive service provider receives full immunity regardless of the specific editing or selection process." 339 F.3d at 1124 (quoted in Doe, 474 F.Supp.2d at 847; Chicago Lawyers' Comm. for Civil Rights Under the Law, Inc. v. Craigslist, Inc., 461 F.Supp.2d 681, 690 n. 7 (N.D.Ill.2006); Dimeo v. Max, 433 F.Supp.2d 523, 530 n. 12 (E.D.Pa.2006); Prickett v. Infousa, Inc., No. 04:05-CV-10, 2006 WL 887431, at *2, 2006 U.S. Dist. LEXIS 21867, at *4 (E.D.Tex. Mar. 30, 2006)).

Significantly, § 230(e) expressly exempts from its scope certain areas of law, such as intellectual property law and federal criminal laws. § 230(e)(1) ("Nothing in this section shall be construed to impair the enforcement of [selected obscenity statutes] or any other Federal criminal statute."); § 230(e)(2) ("Nothing in this section shall be construed to limit or expand any law pertaining to intellectual property."). See also Perfect 10, Inc. v. CCBill LLC, 488 F.3d 1102, 1118 (9th Cir. 2007). Thus, for example, a webhost may still be liable as a publisher or speaker of third-party information that is alleged to infringe a copyright. Notably, the CDA does not exempt the FHA and a host of other federal statutes from its scope. See § 230(e). The FHA existed at the time of the CDA's enactment, yet Congress did not add it to the list of specifically enumerated laws for which publisher and speaker liability was left intact. The absence of a statutory exemption suggests that Congress did not intend to provide special case status to the FHA in connection with immunity under the CDA. See TRW Inc. v. Andrews, 534 U.S. 19, 28, 122 S.Ct. 441, 151 L.Ed.2d 339 (2001) (stating that "[w]here Congress explicitly enumerates certain exceptions to a general prohibition, additional exceptions are not to be implied, in the absence of evidence of a contrary legislative intent.") (citation omitted); see also Craigslist, 519 F.3d at ____ (stating that "[t]he question is not whether Congress gave any thought to the Fair Housing Act, but whether it excluded § 3604(c) from the reach of § 230(c)(1)").

Anticipating the morphing of the Internet and the limits of creative genius and entrepreneurship that fuel its development is virtually impossible. However, Congress explicitly drafted the law to permit this unfettered development of the Internet. Had Congress discovered that, over time, courts across the country have created more expansive immunity than it originally envisioned under the CDA, Congress could have amended the law. But it has not. In fact, just six years ago, Congress approved of the broad immunity that courts have uniformly accorded interactive webhosts under § 230(c).

In 2002, Congress passed the "Dot Kids Implementation and Efficiency Act," which established a new "kids.us" domain for material that is safe for children. Pub.L. No. 107-317, 116 Stat. 2766. Congress stated that the statutory protections of [1188] § 230(c) were extended to certain entities that operated within the new domain. 47 U.S.C. § 941 (stating that certain entities "are deemed to be interactive computer services for purposes of § 230(c)"). The Committee Report that accompanied the statute declared:

The Committee notes that ISPs have successfully defended many lawsuits using section 230(c). The courts have correctly interpreted section 230(c), which was aimed at protecting against liability for such claims as negligence (See, e.g., Doe v. America Online, 783 So.2d 1010 (Fla.2001)) and defamation (Ben Ezra, Weinstein, and Co. v. America Online, 206 F.3d 980 (2000); Zeran v. America Online, 129 F.3d 327 (1997)). The Committee intends these interpretations of section 230(c) to be equally applicable to those entities covered by H.R. 3833.

H.R. REP. No. 107-449, 2002 U.S.C.C.A.N. 1741, 1749 (emphasis added). These statements "reflect the Committee's intent that the existing statutory construction," i.e., broad immunity for interactive webhosts, "be maintained in a new legislative context." Barrett, 146 P.3d at 523 n. 17 (discussing H.R.Rep. No. 107-449); see also Heckler v. Turner, 470 U.S. 184, 209, 105 S.Ct. 1138, 84 L.Ed.2d 138 (1985) (noting that subsequent legislative history can shed useful light on Congressional intent). This express Congressional approval of the courts' interpretation of § 230(c)(1), six years after its enactment, advises us to stay the course of "robust" webhost immunity.

The consequences of the majority's interpretation are far-reaching. Its position will chill speech on the Internet and impede "the continued development of the Internet and other interactive computer services and other interactive media." § 230(b)(1). To the extent the majority strips immunity because of sorting, channeling, and categorizing functions, it guts the heart of § 230(c)(1) immunity. Countless websites operate just like Roommate: they organize information provided by their users into a standardized format, and provide structured searches to help users find information. These sites, and their attendant display, search, and inquiry tools, are an indispensable part of the Internet tool box. Putting a lid on the sorting and searching functions of interactive websites stifles the core of their services.

To the extent the majority strips immunity because the information or query may be illegal under some statute or federal law, this circumstance puts the webhost in the role of a policeman for the laws of the fifty states and the federal system. There are not enough Net Nannies in cyberspace to implement this restriction, and the burden of filtering content would be unfathomable.

To the extent the majority strips immunity because a site solicits or actively encourages content, the result is a direct restriction on the free exchange of ideas and information on the Internet. As noted in the amici curiae brief of the news organizations, online news organization routinely solicit third-party information. Were the websites to face host liability for this content, they "would have no choice but to severely limit its use" and "[s]heer economics would dictate that vast quantities of valuable information be eliminated from websites." Brief of Amici Curiae News Organizations in Support of Roommate.com, LLC 22.

To the extent the majority strips immunity because a website "materially contributed" to the content or output of a website by "specialization" of content, this approach would essentially swallow the immunity provision. The combination of solicitation, sorting, and potential for liability would put virtually every interactive website in this category. Having a website directed to Christians, Muslims, gays, disabled [1189] veterans, or childless couples could land the website provider in hot water.^[54]

Because the statute itself is cumbersome to interpret in light of today's Internet architecture, and because the decision today will ripple through the billions of web pages already online, and the countless pages to come in the future, I would take a cautious, careful, and precise approach to the restriction of immunity, not the broad swath cut by the majority. I respectfully dissent and would affirm the district court's judgment that Roommate is entitled to immunity under § 230(c)(1) of the CDA, subject to examination of whether the bare inquiry itself is unlawful.

[1] This appeal is taken from the district court's order granting defendant's motion for summary judgment, so we view contested facts in the light most favorable to plaintiffs. See Winterrowd v. Nelson, 480 F.3d 1181, 1183 n. 3 (9th Cir.2007).

[2] For unknown reasons, the company goes by the singular name "Roommate.com, LLC" but pluralizes its website's URL, www. roommates.com.

[3] In the online context, "posting" refers to providing material that can be viewed by other users, much as one "posts" notices on a physical bulletin board.

[4] The Fair Housing Act prohibits certain forms of discrimination on the basis of "race, color, religion, sex, familial status, or national origin." 42 U.S.C. § 3604(c). The California fair housing law prohibits discrimination on the basis of "sexual orientation, marital status, . . . ancestry, . . . source of income, or disability," in addition to reiterating the federally protected classifications. Cal. Gov. Code § 12955.

[5] The Supreme Court held some portions of the CDA to be unconstitutional. See Reno v. ACLU, 521 U.S. 844, 117 S.Ct. 2329, 138 L.Ed.2d 874 (1997). The portions relevant to this case are still in force.

[6] Section 230 defines an "interactive computer service" as "any information service, system, or access software provider that provides or enables computer access by multiple users to a computer server." 47 U.S.C. § 230(f)(2); see Carafano v. Metrosplash.com, Inc., 207 F.Supp.2d 1055, 1065-66 (C.D.Cal. 2002) (an online dating website is an "interactive computer service" under the CDA), aff'd, 339 F.3d 1119 (9th Cir.2003). Today, the most common interactive computer services are websites. Councils do not dispute that Roommate's website is an interactive computer service.

[7] The Act also gives immunity to users of third-party content. This case does not involve any claims against users so we omit all references to user immunity when quoting and analyzing the statutory text.

[8] See, e.g., Anthony v. Yahoo! Inc., 421 F.Supp.2d 1257, 1262-63 (N.D.Cal.2006) (Yahoo! is not immune under the CDA for allegedly creating fake profiles on its own dating website).

[9] Prodigy was an online service provider with 2 million users, which seemed like a lot at the time.

[10] A "message board" is a system of online discussion allowing users to "post" messages. Messages are organized by topic — such as the "finance" message board at issue in Stratton Oakmont — and the system generally allows users to read and reply to messages posted by others.

[11] CompuServe was a competing online service provider of the day.

[12] While the Conference Report refers to this as "[o]ne of the specific purposes" of section 230, it seems to be the principal or perhaps the only purpose. The report doesn't describe any other purposes, beyond supporting "the important federal policy of empowering parents to determine the content of communications their children receive through interactive computer services." H.R.Rep. No. 104-458, at 194 (1996) (Conf.Rep.), as reprinted in 1996 U.S.C.C.A.N. 10, 207-08.

[13] The Fair Housing Act prohibits any "statement . . . with respect to the sale or rental of a dwelling that indicates . . . an intention to make [a] preference, limitation, or discrimination" on the basis of a protected category. 42 U.S.C. § 3604(c) (emphasis added). California law prohibits "any written or oral inquiry concerning the" protected status of a housing seeker. Cal. Gov.Code § 12955(b).

[14] The Seventh Circuit has expressly held that inquiring into the race and family status of housing applicants is unlawful. See, e.g., Jancik v. HUD, 44 F.3d 553, 557 (7th Cir.1995).

[15] The dissent stresses the importance of the Internet to modern life and commerce, Dissent at 1176, and we, of course, agree: The' Internet is no longer a fragile new means of communication that could easily be smothered in the cradle by overzealous enforcement of laws and regulations applicable to brick-and-mortar businesses. Rather, it has become a dominant — perhaps the preeminent — means through which commerce is conducted. And its vast reach into the lives of millions is exactly why we must be careful not to exceed the scope of the immunity provided by Congress and thus give online businesses an unfair advantage over their real-world counterparts, which must comply with laws of general applicability.

[16] Roommate argues that Councils waived the argument that the questionnaire violated the FHA by failing to properly raise it in the district court. But, under our liberal pleading standard, it was sufficient for Councils in their First Amended Complaint to allege that Roommate "encourages" subscribers to state discriminatory preferences. See Johnson v. Barker, 799 F.2d 1396, 1401 (9th Cir.1986).

[17] A drop-down menu allows a subscriber to select answers only from among options provided by the website.

[18] See also discussion of Batzel pp. 1170-71 infra.

[19] The dissent may be laboring under a misapprehension as to how the Roommate website is alleged to operate. For example, the dissent spends some time explaining that certain portions of the user profile application are voluntary. Dissent at 1180-82. We do not discuss these because plaintiffs do not base their claims on the voluntary portions of the application, except the "Additional Comments" portion, discussed below, see pp. 1173-75 infra. The dissent also soft-pedals Roommate's influence on the mandatory portions of the applications by referring to it with such words as "encourage" or "encouragement" or "solicitation." Dissent at 1185; see id. at 1188. Roommate, of course, does much more than encourage or solicit; it forces users to answer certain questions and thereby provide information that other clients can use to discriminate unlawfully.

[20] When a prospective subscriber submits his application, Roommate's server presumably checks it to ensure that all required fields are complete, and that any credit card information is not fraudulent or erroneous. Moreover, some algorithm developed by Roommate then decodes the input, transforms it into a profile page and notifies other subscribers of a new applicant or individual offering housing matching their preferences.

[21] Other circuits have held that it is unlawful for housing intermediaries to "screen" prospective housing applicants on the basis of race, even if the preferences arise with landlords. See Jeanty v. McKey & Poague, Inc., 496 F.2d 1119, 1120-21 (7th Cir.1974).

[22] Development, it will be recalled, has many meanings, which differ materially depending on context. Thus, "development" when used as part of the phrase "research and development" means something quite different than when referring to "mental development," and something else again when referring to "real estate development," "musical development" or "economic development."

[23] It is perfectly legal to discriminate along those lines in dating, and thus there can be no claim based solely on the content of these questions.

[24] Requiring website owners to refrain from taking affirmative acts that are unlawful does not strike us as an undue burden. These are, after all, businesses that are being held responsible only for their own conduct; there is no vicarious liability for the misconduct of their customers. Compliance with laws of general applicability seems like an entirely justified burden for all businesses, whether they operate online or through quaint brick-and-mortar facilities. Insofar, however, as a plaintiff would bring a claim under state or federal law based on a website operator's passive acquiescence in the misconduct of its users, the website operator would likely be entitled to CDA immunity. This is true even if the users committed their misconduct using electronic tools of general applicability provided by the website operator.

[25] Of course, the logic of Roommate's argument is not limited to discrimination based on these particular criteria. If Roommate were free to discriminate in providing housing services based on sex, there is no reason another website could not discriminate based on race, religion or national origin. Nor is its logic limited to housing; it would apply equally to websites providing employment or educational opportunities — or anything else, for that matter.

[26] The dissent argues that Roommate is not liable because the decision to discriminate on these grounds does not originate with Roommate; instead, "users have chosen to select characteristics that they find desirable." Dissent at 1185. But, it is Roommate that forces users to express a preference and Roommate that forces users to disclose the information that can form the basis of discrimination by others. Thus, Roommate makes discrimination both possible and respectable.

[27] Apparently, it was common practice for this editor to receive and forward tips from his subscribers. In effect, the newsletter served as a heavily moderated discussion list.

[28] As an initial matter, the Batzel panel held that the defendant newsletter editor was a "user" of an interactive computer service within the definition provided by section 230. While we have our doubts, we express no view on this issue because it is not presented to us. See p. 1162 n. 7 supra. Thus, we assume that the editor fell within the scope of section 230's coverage without endorsing Batzel's analysis on this point.

[29] As Batzel pointed out, there can be no meaningful difference between an editor starting with a default rule of publishing all submissions and then manually selecting material to be removed from publication, and a default rule of publishing no submissions and manually selecting material to be published — they are flip sides of precisely the same coin. Batzel, 333 F.3d at 1032 ("The scope of [section 230] immunity cannot turn on whether the publisher approaches the selection process as one of inclusion or removal, as the difference is one of method or degree, not substance.").

[30] The dissent scores a debater's point by noting that the same activity might amount to "development" or not, depending on whether it contributes materially to the illegality of the content. Dissent at 3489. But we are not defining "development" for all purposes; we are defining the term only for purposes of determining whether the defendant is entitled to immunity for a particular act. This definition does not depend on finding substantive liability, but merely requires analyzing the context in which a claim is brought. A finding that a defendant is not immune is quite distinct from finding liability: On remand, Roommate may still assert other defenses to liability under the Fair Housing Act, or argue that its actions, do not violate the Fair Housing Act at all. Our holding is limited to a determination that the CDA provides no immunity to Roommate's actions in soliciting and developing the content of its website; whether that content is in fact illegal is a question we leave to the district court.

[31] We disavow any suggestion that Carafano holds an information content provider automatically immune so long as the content originated with another information content provider. 339 F.3d at 1125.

[32] Section 230 requires us to scrutinize particularly closely any claim that can be boiled down to the failure of an interactive computer service to edit or block user-generated content that it believes was tendered for posting online, see pp. 1170-71 supra, as that is the very activity Congress sought to immunize by passing the section. See pp. 1162-64 supra.

[33] The dissent coyly suggests that our opinion "sets us apart from" other circuits, Dissent at 1177, 1179-80, carefully avoiding the phrase "intercircuit conflict." And with good reason: No other circuit has considered a case like ours and none has a case that even arguably conflicts with our holding today. No case cited by the dissent involves active participation by the defendant in the creation or development of the allegedly unlawful content; in each, the interactive computer service provider passively relayed content generated by third parties, just as in Stratton Oakmont, and did not design its system around the dissemination of unlawful content.

In Chicago Lawyers' Committee for Civil Rights Under Law, Inc. v. craigslist, Inc., 519 F.3d 666 (7th Cir.2008), the Seventh Circuit held the online classified website craigslist immune from liability for discriminatory housing advertisements submitted by users. Craigslist's service works very much like the "Additional Comments" section of Roommate's website, in that users are given an open text prompt in which to enter any description of the rental property without any structure imposed on their content or any requirement to enter discriminatory information: "Nothing in the service craigslist offers induces anyone to post any particular listing or express a preference for discrimination. . . ." 519 F.3d at ____. We similarly hold the "Additional Comments" section of Roommate's site immune, see pp. 1173-75 infra. Consistent with our opinion, the Seventh Circuit explained the limited scope of section 230(c) immunity. Craigslist, at ___ _ ___. More directly, the Seventh Circuit noted in dicta that "causing a particular statement to be made, or perhaps [causing] the discriminatory content of a statement" might be sufficient to create liability for a website. At ____ (emphasis added). Despite the dissent's attempt to imply the contrary, the Seventh Circuit's opinion is actually in line with our own. In Universal Communication Systems v. Lycos, Inc., the First Circuit held a message board owner immune under the CDA for defamatory comments posted on a message board. 478 F.3d 413 (1st Cir.2007). The allegedly defamatory comments were made without any prompting or encouragement by defendant: "[T]here is not even a colorable argument that any misinformation was prompted by Lycos's registration process or its link structure." Id. at 420.

Green v. America Online, 318 F.3d 465 (3d Cir.2003), falls yet farther from the mark. There, AOL was held immune for derogatory comments and malicious software transmitted by other defendants through AOL's "Romance over 30" "chat room." There was no allegation that AOL solicited the content, encouraged users to post harmful content or otherwise had any involvement whatsoever with the harmful content, other than through providing "chat rooms" for general use.

In Ben Ezra, Weinstein, and Co. v. America Online Inc., 206 F.3d 980 (10th Cir.2000), the Tenth Circuit held AOL immune for relaying inaccurate stock price information it received from other vendors. While AOL undoubtedly participated in the decision to make stock quotations available to members, it did not cause the errors in the stock data, nor did it encourage or solicit others to provide inaccurate data. AOL was immune because "Plaintiff could not identify any evidence indicating Defendant [AOL] developed or created the stock quotation information." Id. at 985 n. 5. And, finally, in Zeran v. America Online, Inc., 129 F.3d 327 (4th Cir.1997), the Fourth Circuit held AOL immune for yet another set of defamatory and harassing message board postings. Again, AOL did not solicit the harassing content, did not encourage others to post it, and had nothing to do with its creation other than through AOL's role as the provider of a generic message board for general discussions.

[34] "The female we are looking for hopefully wont [sic] mind having a little sexual incounter [sic] with my boyfriend and I [very sic]."

[35] "We are 3 Christian females who Love our Lord Jesus Christ. . . . We have weekly bible studies and bi-weekly times of fellowship."

[36] It is unclear whether Roommate performs any filtering for obscenity or "spam," but even if it were to perform this kind of minor editing and selection, the outcome would not change. See Batzel, 333 F.3d at 1031.

[37] Nor would Roommate be the developer of discriminatory content if it provided a free-text search that enabled users to find keywords in the "Additional Comments" of others, even if users utilized it to search for discriminatory keywords. Providing neutral tools for navigating websites is fully protected by CDA immunity, absent substantial affirmative conduct on the part of the website creator promoting the use of such tools for unlawful purposes.

[38] It's true that, under a pedantic interpretation of the term "develop," any action by the website — including the mere act of making a text box available to write in — could be seen as "develop[ing]" content. However, we have already rejected such a broad reading of the term "develop" because it would defeat the purpose of section 230. See pp. 1167-69 supra.

[39] The dissent also accuses us of creating uncertainty that will chill the continued growth of commerce on the Internet. Dissent at 1186-87. Even looking beyond the fact that the Internet has outgrown its swaddling clothes and no longer needs to be so gently coddled, see p. 1164-65 n. 15 supra, some degree of uncertainty is inevitable at the edge of any rule of law. Any immunity provision, including section 230, has its limits and there will always be close cases. Our opinion extensively clarifies where that edge lies, and gives far more guidance than our previous cases. While the dissent disagrees about the scope of the immunity, there can be little doubt that website operators today know more about how to conform their conduct to the law than they did yesterday.

However, a larger point remains about the scope of immunity provisions. It's no surprise that defendants want to extend immunity as broadly as possible. We have long dealt with immunity in different, and arguably far more important, contexts — such as qualified immunity for police officers in the line of duty, see Clement v. City of Glendale, 518 F.3d 1090 (9th Cir.2008) — and observed many defendants argue that the risk of getting a close case wrong is a justification for broader immunity. Accepting such an argument would inevitably lead to an endless broadening of immunity, as every new holding creates its own borderline cases.

[40] We do not address Roommate's claim that its activities are protected by the First Amendment. The district court based its decision entirely on the CDA and we refrain from deciding an issue that the district court has not had the opportunity to evaluate. See Mukherjee v. INS, 793 F.2d 1006, 1010 (9th Cir.1986).

[41] Internet World Stats, World Internet Users: December 2007, http:// www.internet worldstats.com/stats.htm (last visited Mar. 14, 2008); Netcraft, February 2008 Web Server Survey, http://news.rietcraft.com/ archives/web — server — survey.html (last visited Mar. 14, 2008).

[42] Ashcroft v. ACLU, 535 U.S. 564, 566, 122 S.Ct. 1700, 152 L.Ed.2d 771 (2002).

[43] Carafano v. Metrosplash.com, Inc., 339 F.3d 1119, 1123 (9th Cir.2003).

[44] According to one commentator, in 1985, there were approximately 1,000 host computers connected to the Internet; by 1995, that number had exploded to 4,000,000. Paul H. Arne, New Wine in Old Bottles: The Developing Law of the Internet, 416 PLI/Pat 9, 15 (Sept.1995).

[45] The complaint centers on the responses and profiles generated by the users. To the extent that the inquiry in isolation is part of the claims, then I agree with Part 1 of the majority's opinion that § 230(c)(1) would not protect Roommate. However, I cannot join the majority insofar as it eviscerates the distinction between traditional publishers and webhosts. See, e.g., Maj. Op. at 1164 (ignoring the Congressional carveout for interactive service providers and concluding that if a face-to-face transaction were illegal, it could not be legal in cyberspace).

[46] The statute also seeks to "remove disincentives for the development and utilization of blocking and filtering technologies" and "to ensure vigorous enforcement of Federal criminal laws to deter and punish trafficking in obscenity, stalking, and harassment by means of computer." § 230(b)(4), (5).

[47] The second part of this subsection, § 230(c)(2), is more accurately characterized as an immunity provision, but is not relevant to our discussion here. Compare 47 U.S.C. § 230(c)(2) (stating that "[n]o provider or user of an interactive computer service shall be held liable . . .") (emphasis added).

[48] A user who is a room-seeker fills out an equivalent section named "About Me."

[49] The following is an example of a member profile:

The Basics

Rent: $800 per month + $800 deposit

Lease: 6 month

Date available: 09/01/04 (14 days)

Utilities included: N/A

Features: Private bedroom, Private bathroom

Residence & Vicinity

Building: House, 2 bed, 1.5 bath

Features: N/A

Location: (Central) Long Beach, CA

Household

Occupant: 1, Age 26, Male (straight)

Occupation: Student

Smoking habits: Outside smoker

Cleanliness: About average

Children: Children will not be living with us

Pets: Dog(s)

Preferences

Age group: 18-99

Gender: Male (straight or gay), Female (straight or lesbian)

Smoking: Smoking okay

Cleanliness level. Clean, Average, Messy

Pets: Dog okay. Cat okay, Caged pet okay

Children: Children okay

Comments

LOOKING FOR CHILL ROOMATE [sic] TO SHARE 2 BR

HOUSE WITH DOG AND FERRET-RENT 800/MO + utill.6mo.lease.

[50] Congress also stated in the CDA that "[i]t is the policy of the United States to — (1) to promote the continued development of the Internet and other interactive computer services and other interactive media," and "(4) to remove disincentives for the development and utilization of blocking and filtering technologies . . ." § 230(b)(1), (4) (emphasis added).

[51] The majority's notion of using a different definition of "development" digs the majority into a deeper hole. See Maj. Op. at 1167-69. For example, adopting the Wikipedia definition of "content development" — "the process of researching, writing, gathering, organizing and editing information for publication on web sites" — would run us smack into the sphere of Congressionally conferred immunity. Wikipedia, Content Development (Web), http://en.wikipedia.org/w/index.php?title= Content — development — web. & oldid = XXXXXXXXX (last visited Mar. 24, 2008). Both our circuit and others have steadfastly maintained that activities such as organizing or editing information are traditional editorial functions that fall within the scope of CDA immunity. See, e.g., Carafano, 339 F.3d at 1124-25; Zeran, 129 F.3d at 330. Likewise, an alternative definition of "development" from Webster's such as "a making usable or available" sweeps too broadly, as "making usable or available" is precisely what Google and Craigslist do. In an effort to cabin the reach of the opinion, the majority again goes back to whether the content is legal, i.e., a dating website that requires sex, race, religion, or marital status is legal because it is legal to discriminate in dating. See Maj. Op. at 1169. Of course this approach ignores whether the claim may be one in tort, such as defamation, rather than a statutory discrimination claim. And, this circularity also circumvents the plain language of the statute. Interestingly, the majority has no problem offering up potentially suitable definitions of "development" by turning to dictionaries, but it fails to explain why, and from where, it plucked its definition of "development" as "materially contributing to [the] alleged unlawfulness" of content. See Maj. Op. at 1168.

[52] Again, Roommate does not force users to disclose preferences as to roommate characteristics.

[53] The First Circuit has noted that "[i]t is not at all clear that there is a culpable assistance exception to Section 230 immunity[,]" similar to the notion of secondary liability under the Electronic Communications Privacy Act of 1986. Universal Commc'n, 478 F.3d at 421. But it also stated that it "need not decide whether a claim premised on active inducement might be consistent with Section 230 in the absence of a specific exception." Id.

[54] It is no surprise that there are countless specialized roommate sites. See, e.g., http:// islam.to/housing/index.php, http://christian-roommates.com, and http://prideroommates. com.