Interest of Amicus[1]

The International Center for Law & Economics (“ICLE”) is a nonprofit, non-partisan global research and policy center that builds intellectual foundations for sensible, economically grounded policy. ICLE promotes the use of law and economics methodologies and economic learning to inform policy debates and has longstanding expertise evaluating law and policy.

ICLE has an interest in ensuring that First Amendment law promotes the public interest by remaining grounded in sensible rules informed by sound economic analysis. ICLE scholars have written extensively on issues related to social media regulation and free speech. See, e.g., Geoffrey A. Manne, Ben Sperry, & Kristian Stout, Who Moderates the Moderators?: A Law & Economics Approach to Holding Online Platforms Accountable Without Destroying the Internet, 49 Rutgers Computer & Tech. L. J. 26 (2022); Ben Sperry, Knowledge and Decisions in the Information Age: The Law & Economics of Regulating Misinformation on Social-Media Platforms, 59 Gonzaga L. Rev., forthcoming (2023); Br. of Internet Law Scholars, Gonzalez v. Google; Jamie Whyte, Polluting Words: Is There a Coasean Case to Regulate Offensive Speech?, ICLE White Paper (Sep. 2021); Ben Sperry, An L&E Defense of the First Amendment’s Protection of Private Ordering, Truth on the Market (Apr. 23, 2021); Liability for User-Generated Content Online: Principles for Lawmakers (Jul. 11, 2019).

Statement

The pair of NetChoice cases before the Court presents the opportunity to bolster the Court’s longstanding jurisprudence on state action and editorial discretion by affirming that the First Amendment applies to Internet speech without disfavor. See Reno v. ACLU, 521 U.S. 844, 870 (1997) (finding “no basis for qualifying the level of First Amendment scrutiny that should be applied” to the Internet).

The First Amendment protects social media companies’ rights to exercise their own content moderation policies free from government interference. Social media companies are private actors with the same right to editorial discretion over disseminating third-party speech as offline equivalents like newspapers and cable operators. See Manhattan Cmty. Access Corp. v. Halleck, 139 S. Ct. 1921, 1926 (2019); Mia. Herald Publ’g Co. v. Tornillo, 418 U.S. 241 (1974); Turner Broad. Sys. v. FCC, 512 U.S. 622 (1994).

Consistent with that jurisprudence, the Court should conclude that social media companies are private actors fully capable of taking part in the marketplace of ideas through their exercise of editorial discretion, free from government interference.

Summary of Argument

“The most basic of all decisions is who shall decide.” Thomas Sowell, Knowledge and Decisions 40 (2d ed. 1996). Under the First Amendment, the general rule is that private actors get to decide what speech is acceptable. It is not the government’s place to censor speech or to require private actors to open their property to unwanted speech. The market process determines speech rules on social media platforms[2] just as it does in the offline world.

The animating principle of the First Amendment is to protect this “marketplace of ideas.” “The theory of our Constitution is ‘that the best test of truth is the power of the thought to get itself accepted in the competition of the market.’” United States v. Alvarez, 567 U.S. 709, 728 (2012) (quoting Abrams v. United States, 250 U.S. 616, 630 (1919) (Holmes, J., dissenting)). To facilitate that competition, the Constitution staunchly protects the liberty of private actors to determine what speech is acceptable, largely free from government regulation of this marketplace. See Halleck, 139 S. Ct. at 1926 (“The Free Speech Clause of the First Amendment constrains governmental actors and protects private actors….”).

Importantly, one way private actors participate in the marketplace of ideas is through private ordering—by setting speech policies for their own private property, enforceable by common law remedies under contract and property law. See id. at 1930 (a “private entity may thus exercise editorial discretion over the speech and speakers in the forum”).

Protecting private ordering is particularly important with social media. While the challenged laws concern producers of social media content, producers are only a sliver of social media users. The vast majority of social media users are content consumers, and it is for their benefit that social media companies moderate content. Speech, even when lawful and otherwise protected by the First Amendment, can still be harmful, at least from the point of view of listeners. Social media companies must balance users’ demand for speech with the fact that not everyone wants to consume every possible type of speech.

The issue is how best to optimize the benefits of speech while minimizing negative speech externalities. Speech produced on social media platforms causes negative externalities when some consumers are exposed to speech they find offensive, disconcerting, or otherwise harmful. Those consumers may stop using the platform as a result. On the other hand, if limits on speech production are too extreme, speech producers and consumers may seek other speech platforms.

To optimize the value of their platforms, social media companies must consider how best to keep users—both producers and consumers of speech—engaged. Major social media platforms mainly generate revenue through advertisements. This means a loss in user engagement could reduce the value to advertisers, and thus result in less advertising revenue. In particular, a loss in engagement by high-value users could result in less advertising, and that in turn, diminishes incentives to invest in the platform. Optimizing a platform requires satisfying users who are valuable to advertisers.

Major social media platforms have developed moderation policies in response to market demand to protect their users from speech those users consider harmful. This editorial control is protected First Amendment activity.

On the other hand, the common carriage justifications Texas and Florida offer for their restrictions on social media platforms’ control over their own property do not save the States’ impermissible intervention into the marketplace of ideas. Two of the most prominent legal justifications for common carriage regulation—holding one’s property open to all-comers and market power—do not apply to social media companies. Major social media companies require all users to accept terms of service, which limit what speech is allowed. And assuming market power can justify common carriage, neither Florida nor Texas even attempted to make such a finding, making at best mere assertions.

The States’ intervention is more like treating social media platforms as company towns—an outdated approach that this Court should reject as inconsistent with First Amendment doctrine and utterly unsuitable to the Internet Age.

Argument

I. Social Media Platforms Are Best Positioned to Optimize Their Platforms To Serve Their Users’ Speech Preferences.

The First Amendment promotes a marketplace of ideas. To have a marketplace of any kind, there must be strong private property rights and enforceable contracts that enable entrepreneurs to discover the best ways to serve consumers. See generally Hernando de Soto, The Mystery of Capital (2000). As full participants in the marketplace of ideas, social media platforms must be free to exercise their own editorial policies and have choice over which ideas they allow on their platforms. Otherwise, there is no marketplace of ideas at all, but either a government-mandated free-for-all where voices struggle to be heard or an overly restricted forum where the government censors disfavored ideas.

The marketplace analogy is apt when considering First Amendment principles because, like virtually any other human activity, speech has both benefits and costs. Like other profit-driven market endeavors, it is ultimately the subjective, individual preferences of consumers that determine how to manage those tradeoffs. The nature of what is deemed offensive is obviously context- and listener-dependent, but the parties best suited to set and enforce appropriate speech rules are the property owners subject to the constraints of the marketplace.

When it comes to speech, an individual’s desire for an audience must be balanced with a prospective audience’s willingness to listen. Formal economic institutions acting in the marketplace must strike the proper balance between these desires and have an incentive to get it right or they could lose consumers. Asking government to make categorical decisions for all of society is substituting centralized evaluation of the costs and benefits of access to communications for the individual decisions of many actors, including property owners who open their property to third party speech. As the economist Thomas Sowell put it, “that different costs and benefits must be balanced does not in itself imply who must balance them?or even that there must be a single balance for all, or a unitary viewpoint (one ‘we’) from which the issue is categorically resolved.” Thomas Sowell, Knowledge and Decisions 240 (2d ed. 1996).

Rather than incremental decisions on how and under what terms individuals may relate to one another on a particular platform—which can evolve over time in response to changes in what individuals find acceptable—governments can only hand down categorical guidelines through precedential decisions: “you must allow a, b, and c speech” or “you must not allow x, y, and z speech.”

This freedom to experiment and evolve is vital in the social-media sphere, where norms about speech are in constant flux. Social media users often impose negative externalities on other users through their speech. Thus, social media companies must resolve social-cost problems among their users by balancing their speech interests.

In his famous work “The Problem of Social Cost,” the economist Ronald Coase argued that the traditional approach to regulating externalities was misguided because it overlooked the reciprocal nature of harms. Ronald H. Coase, The Problem of Social Cost, 3 J. L. & Econ. 1, 2 (1960). For example, the noise from a factory is a potential cost to the doctor next door who consequently cannot use his office to conduct certain testing, and simultaneously the doctor moving his office next door is a potential cost to the factory’s ability to use its equipment. In a world of well-defined property rights and low transaction costs, the initial allocation of a right would not matter, because the parties could bargain to overcome the harm in a beneficial manner—i.e., the factory could pay the doctor for lost income or to set up sound-proof walls, or the doctor could pay the factory to reduce the sound of its machines. But in the real world, where there are often significant transaction costs, who has the initial right matters because it is unlikely that the right will get to the highest valued use.

Similarly, on social media, speech that some users find offensive or false may be inoffensive or even patently true to other users. Protecting one group from offensive speech necessarily imposes costs on the group that favors the same speech. There is a reciprocal nature to the harms of speech, much as with other forms of nuisance. Due to transaction costs, it is unlikely that users will be able to effectively bargain to a solution on speech harms. There is a significant difference, though. Unlike the situation of the factory owner and the doctor, social media users are all using the property of social media companies. And those companies are best positioned to—and must be allowed to—balance these varied interests in real-time to optimize their platform’s value in response to consumer demand.

Social media companies are what economists call “multi-sided” platforms. See generally David S. Evans & Richard Shmalensee, Matchmakers: The New Economics of Multisided Platforms (2016). They are for-profit businesses, and the way they generate profits is by acting as intermediaries between users and advertisers. If they fail to serve their users well, those users will abandon the platform. Without users, advertisers would have no interest in buying ads. And without advertisers, there is no profit to be made.

As in any other community, “[i]nteractions on multi-sided platforms can involve behavior that some users find offensive.” David S. Evans, Governing Bad Behavior by Users of Multi-Sided Platforms, 27 Berkeley Tech. L.J. 1201, 1215 (2012). As a result, “[p]eople may incur costs [from] unwanted exposure to hate speech, pornography, violent images, and other offensive content.” Id. And “[e]ven if they are not exposed to this content, they may dislike being part of a community in which such behavior takes place.” Id.

These cases challenge laws that cater to one set of social media users—producers of speech on social media platforms. But social media platforms must be at least as sensitive to their speech consumers. Indeed, the one-percent rule—“a vast majority of user-generated content in any specific community comes from the top 1% of active users”^[3]—teaches that speech-consuming users may be even more important because they far outnumber producers. In turn, less intense users are usually the first to leave a platform, and their exit may cascade into total platform collapse. See, e.g., János Török & János Kertész, Cascading Collapse of Online Social Networks, 7 Sci. Rep., art. 16743 (2017).

Social media companies thus need to optimize the value of their platform by setting rules that keep users—mostly speech consumers—sufficiently engaged that there are advertisers who will pay to reach them. Even more, social media platforms must encourage engagement by the right users. To attract advertisers, platforms must ensure individuals likely to engage with advertisements remain active on the platform.^[4] Platforms ensure this optimization by setting and enforcing community rules.

In addition, like users, advertisers themselves have preferences social media platforms must take into account. Advertisers may threaten to pull ads if they do not like the platform’s speech-governance decisions. For instance, after Elon Musk restored the accounts of Twitter users who had been banned by the company’s prior leadership, major advertisers left the platform. See Kate Conger, Tiffany Hsu, & Ryan Mac, Elon Musk’s Twitter Faces Exodus of Advertisers and Executives, N.Y. Times (Nov. 1, 2022); Ryan Mac & Tiffany Hsu, Twitter’s US Ad Sales Plunge 59% as Woes Continue, N.Y. Times (Jun. 5, 2013).

Thus, it is no surprise that in the cases of major social media companies, the platforms have set content-moderation standards that restrict many kinds of speech. See generally Kate Klonick, The New Governors: The People, Rules, and Processes Governing Online Speech, 131 Harv. L. Rev. 1598 (2018).

The bottom line is that the market process leaves the platforms themselves best positioned to make these incremental editorial decisions about their users’ preferences on speech, in response to the feedback loop between consumer, producer, and advertiser demand. It should go without saying that social media users do not necessarily want more opportunities to say and hear certain speech. Forcing social media companies to favor one set of users—a fraction of speech producers—by forbidding “viewpoint discrimination” favored by other users is unwarranted and unlawful interference in those companies’ editorial discretion. That interference threatens rather than promotes the marketplace of ideas.

II. The First Amendment Protects Private Ordering of Speech, Including Social Media Platform Moderation Polices.

The First Amendment protects the right of social media platforms to serve the speech preferences of their users through their moderation policies.

The “text and original meaning [of the First and Fourteenth Amendments], as well as this Court’s longstanding precedents, establish that the Free Speech Clause prohibits only governmental abridgment of speech. The Free Speech Clause does not prohibit private abridgment of speech.” Halleck, 139 S. Ct. at 1928. The First Amendment’s reach does not grow when private property owners open their property for speech. If such property owners were “subject to First Amendment constraints” and thus “lose the ability to exercise what they deem to be appropriate editorial discretion within that open forum” they would “face the unappetizing choice of allowing all comers or closing the platform altogether.” Id. at 1930. That is, the First Amendment respects—indeed protects—private ordering.

So, while the First Amendment protects the right of individuals to speak (and receive speech) without fear of legal repercussions in most instances, it does not make speech consequence-free, nor does it mandate the carrying of all speech in private spaces.

“Bad” speech has, in fact, long been kept in check via informal means, or what one might call “private ordering.” In this sense, property rights and contract law have long played a crucial role in determining the speech rules of any given space.

For instance, a man would be well within his legal rights to eject a guest from his home for using racial epithets. As a property owner, he would not only have the right to ask that person to leave but could exercise his right to eject that person as a trespasser—if necessary, calling the police to assist him. Similarly, one could not expect to go to a restaurant and yell at the top of her lungs about political issues and expect the venue to abide. A bar hosting an “open mic night” and thus opening itself up to speech is still within its rights to end a performance so offensive it could lead to a loss of patrons. Subject to narrow exceptions, property owners determine acceptable speech on their property and may enforce those rules by excluding those who refuse to comply.

A. Social media platforms are not state actors.

One exception to this strong distinction between state and private action is when a “private entity performs a traditional, exclusive public function.” See Halleck, 139 S. Ct. at 1928. In those cases, there may be a right to free speech that operates against a private actor. See Marsh v. Alabama, 326 U.S. 501 (1946).

Proceeding from Marsh, many litigants seize upon this Court’s recent analogizing social media to the “modern public square.” Packingham v. N. Carolina, 137 S. Ct. 1730, 1737 (2017). They argue social media companies are like a company town or town square and so lack the discretion to restrict speech protected by the First Amendment. But cases since Marsh make clear that the state-actor exception is exceptionally narrow.

In Marsh, this Court found that a company town, while private, was a state actor for purposes of the First Amendment. At issue was whether the company town could prevent a Jehovah’s Witness from passing out literature on the town’s sidewalks. The Court noted that “[o]wnership does not always mean absolute dominion. The more an owner, for his advantage, opens up his property for use by the public in general, the more do his rights become circumscribed by the statutory and constitutional rights of those who use it.” Marsh, 326 U.S. at 506. The Court proceeded to balance private property rights with First Amendment rights, determining that, in company towns, the First Amendment’s protections should be in the “preferred position.” See id. at 509.

The Court later extended this finding to shopping centers, finding they were the “functional equivalent” to the business district in Marsh, and thus finding that a shopping center could not restrict peaceful picketing of a grocery story by a local food-workers union. Food Employees v. Logan Valley Plaza, 391 U.S. 308, 318, 325 (1968).

But the Court began retreating from both Logan Valley and Marsh just a few years later in Lloyd Corp. v. Tanner, 407 U.S. 551 (1972), which concerned hand-billing in a shopping mall. Noting the “economic anomaly” that was company towns, the Court said Marsh “simply held that where private interests were substituting for and performing the customary functions of government, First Amendment freedoms could not be denied where exercised in the customary manner on the town’s sidewalks and streets.” Id. at 562 (emphasis added).

Building on Tanner, the Court went a step further in Hudgens v. NLRB, 424 U.S. 507 (1976), reversing Logan Valley and more severely cabining Marsh. Hudgens involved picketing on private property, and the Court concluded bluntly that, “under the present state of the law the constitutional guarantee of free expression has no part to play in a case such as this[.]” Id. at 521. Marsh is now a narrow exception, the Court explained, limited to situations where private property has taken on all attributes of a town. See id. at 516. And following Hudgens, the Court further limited the public-function test to “the exercise by a private entity of powers traditionally exclusively reserved to the State.” See Jackson v. Metropolitan Edison Co., 419 U.S. 345, 352 (1974).

Today it is well-established that “the constitutional guarantee of free speech is a guarantee only against abridgment by government, federal or state.” Hudgens, 424 U.S. at 513. Purely private actors—even those who open their property to the public—are not subject to First-Amendment limits on how they use their property.

The Court reaffirmed that rule recently in Halleck, which considered whether a public-access channel operated by a cable provider was a state actor. Summarizing the case law, the Court said the test required more than just a finding that the government at some point exercised the same function or that the function serves the public good. Instead, the government must have “traditionally and exclusively performed the function.” Halleck, 139 S. Ct. at 1929 (emphasis in original).

The Court then found that merely operating as a public forum for speech is not a function traditionally and exclusively performed by the government. And because “[it] is not an activity that only governmental entities have traditionally performed,” a private actor providing a forum for speech retains “editorial discretion over the speech and speakers in the forum.” Id. at 1930.

Following this Court’s state-actor jurisprudence, federal courts have consistently found social media companies are not equivalent to company towns and thus not subject to First Amendment constraints. Unlike the company town, where those within their geographical confines have little choice but to deal with them as if they are the government themselves, social media users can simply use alternative means to convey speech or receive it. The Ninth Circuit, for instance, squarely rejected the argument that social media companies fulfill a traditional, public function. See Prager Univ. v. Google, LLC, 951 F.3d 991, 996-99 (9th Cir. 2020). Every federal court to consider whether social media companies are state actors under this theory has found the same. See, e.g., Freedom Watch, Inc. v. Google Inc., 816 F. App’x 497, 499 (D.C. Cir. 2020); Brock v. Zuckerberg, 2021 WL 2650070, at *3 (S.D.N.Y. Jun. 25, 2021); Zimmerman v. Facebook, Inc., 2020 WL 5877863 at *2 (N.D. Cal. Oct. 2, 2020); Ebeid v. Facebook, Inc., 2019 WL 2059662 at *6 (N.D. Cal. May 9, 2019); Green v. YouTube, LLC, 2019 WL 1428890, at *4 (D.N.H. Mar. 13, 2019); Nyabwa v. Facebook, 2018 WL 585467, at *1 (S.D. Tex. Jan. 26, 2018); Shulman v. Facebook.com, 2017 WL 5129885, at *4 (D.N.J. Nov. 6, 2017).

B. Social media companies have a right to editorial discretion.

Private actors have the right to editorial discretion that cannot generally be overcome by state action compelling the dissemination of speech. See Mia. Herald Publ’g Co. v. Tornillo, 418 U.S. 241 (1974); Turner Broad. Sys. v. FCC, 512 U.S. 622 (1994). This is particularly important for private actors whose business is disseminating speech, like newspapers, cable operators, and social media companies.

In Tornillo, the Court struck a right-to-reply statute for political candidates because it “compel[s] editors or publishers to publish that which ‘reason tells them should not be published.’” 418 U.S. at 256. The Court established a general rule that the limits on media companies’ editorial discretion were not defined by government edict but by “the acceptance of a sufficient number of readers—and hence advertisers —to assure financial success; and, second, the journalistic integrity of its editors and publishers.” Id. at 255 (citing Columbia Broadcasting System, Inc. v. Democratic Nat’l Comm., 412 U. S. 94, 117 (1973)). In other words, the limits on how private entities exercise their editorial discretion comes from the marketplace of ideas itself—the preferences of speech consumers, advertisers, and the property owners—not the government.

The size and influence of social media companies does not shrink Tornillo’s effect. No matter how large the editor or the forum, the government still may not coerce private entities to disseminate speech. See id. at 254 (“However much validity may be found in these arguments [about monopoly power], at each point the implementation of a remedy such as an enforceable right of access necessarily calls for some mechanism .?.?.?If it is governmental coercion, this at once brings about a confrontation with the express provisions of the First Amendment.”). Alleged market power is insufficient to justify compelling the dissemination of speech by social media companies.

Turner confirms that market power is irrelevant. There the Court began with “an initial premise: Cable programmers and cable operators engage in and transmit speech, and they are entitled to the protection of the speech and press provisions of the First Amendment.” 512 U.S. at 636. While the Court nonetheless applied intermediate scrutiny, it did so based on technological differences in transmission by newspapers and cable television, and the fact that the law was content-neutral. The level of scrutiny thus turns on “the special characteristics” of transmission, not “the economic characteristics” of the market. Id. at 640.

Returning to Tornillo, the Court reasoned that the law violated the First Amendment by intruding upon the company’s editorial discretion. See 418 U.S. at 258. Like newspapers, social media platforms are “more than a passive receptable for news, comment, and advertising,” as their “choice of material,” their “decisions made as to the limitations on the size and content of the paper” and their “treatment of public issues and public officials—whether fair or unfair—constitute the exercise of editorial control and judgment.” Id. Indeed, that exercise of editorial control and judgment is central to a platform’s retention of speech consumers and attraction of advertisers targeting those users, and thus the platform’s continued survival. See supra, pp. ___.

Accordingly, federal courts rightly have called government actions into question when they violate the right of social media platforms to exercise editorial discretion. See NetChoice, LLC v. Bonta, 2023 WL 6135551, at *15 (N.D. Cal. Sept. 18, 2023); O’Handley v. Padilla, 579 F. Supp. 3d 1163, 1186-88 (N.D. Cal. Jan. 10, 2022); see also Murthy v. Missouri, No. 23-411, 2023 WL 6935337, at *2 (U.S. Oct. 20, 2023) (Alito, J., dissenting) (“The injunction applies only when the Government crosses the line and begins to coerce or control others’ [i.e. the social media companies’] exercise of their free-speech [i.e. editorial discretion] rights.”).

Thus, the Fifth Circuit’s claim in Paxton that “the Supreme Court’s cases do not carve out ‘editorial discretion’ as a special category of First-Amendment-protected expression,” 49 F.4th at 463, is demonstrably wrong. The Court has established that private actors have a right to exercise editorial discretion concerning speech on their property. See Halleck (using the phrase “editorial discretion” 11 times). Social media platforms have the same right.

C. Strict scrutiny applies.

As social media companies have a right to editorial discretion, the next question is the level of scrutiny the challenged statutes must satisfy. Strict scrutiny is proper, because social media platforms are much more like the newspapers in Tornillo than the cable companies in Turner.

In Turner, the Court found:

[The] physical connection between the television set and the cable network gives the cable operator bottleneck, or gatekeeper, control over most (if not all) of the television programming that is channeled into the subscriber’s home .?.?.?. [U]nlike speaker in other media, [cable operators] can thus silence the voice of competing speakers with a mere flick of the switch.

512 U.S. at 656. Social media platforms have no physical control of the connection to the home, and thus no practical ability to exclude competing voices or platforms. The internet architecture simply does not allow them to stop users from using other sites to find speech or speak. Strict scrutiny should apply to SB 7072 and HB 20.

Likewise, compelling social media companies to allow speech contrary to their terms of service is fundamentally different than mandating access for military recruiters in law schools or requiring shopping malls to allow the peaceful exercise of speech in areas held open to the public. Contra Paxton, 49 F.4th at 462-63. In those instances, there was no identification of the venue with the message. See Rumsfeld v. Forum for Acad. & Inst. Rights, Inc., 547 U.S. 47, 65 (2006); PruneYard Shopping Ctr. v. Robins, 447 U.S. 74, 86-88 (1980).

Here, the moderation decisions of social media companies do have implications for advertisers who do not want their brand associated with certain content. See Jonathan Vanian, Apple, Disney, other media companies pause advertising on X after Elon Musk boosted antisemitic tweet, CNBC (Nov. 17, 2023);[5] Caleb Ecarma, Twitter Can’t Seem to Buck Its Advertisers-Don’t-Want-to-Be-Seen-Next-to-Nazis Problem, Vanity Fair (Aug. 17, 2023);[6] Ryan Mac & Tiffany Hsu, Twitter’s US Ad Sales Plunge 59% as Woes Continue, N.Y. Times (Jun. 5, 2023).[7] Similarly, users will exit if they don’t enjoy the experience of the platform. See Steven Vaughan-Nichols, Twitter seeing ‘record user engagement’? The data tells a different story, ZDNet (Jun. 30, 2023).[8] Speech by social media companies disavowing what is said by some users of their platforms does not prevent advertisers and much of the public from identifying user speech with the platform.

Moreover, both the Florida and Texas laws are discriminate based upon content, as a reviewing court would have to consider what speech is at issue to determine whether a social media company can moderate it. This makes the laws different than those at issue in Turner, and offer an alternative reason they should be subject to strict scrutiny.

Section 230 of the Communications Act does not change this analysis. Contra Paxton, 49 F.4th at 465-66. Section 230 supplements the First Amendment’s protection of editorial discretion by granting “providers and users of an interactive computer service” immunity from (most) lawsuits for speech generated by other “information content providers” on their platforms. See 47 U.S.C. §230(c). The animating reason for Section 230 was to provide “protection for private blocking and screening” by preventing lawsuits over third party content that was left up, see Section 230(c)(1), or over third-party content that was taken down, see Section 230(c)(2). See also Geoffrey A. Manne, Ben Sperry, & Kristian Stout, Who Moderates the Moderators?: A Law & Economics Approach to Holding Online Platforms Accountable Without Destroying the Internet, 49 Rutgers Computer & Tech. L. J. 26, 39-41 (2022). Section 230 encourages social media companies to use their underlying First Amendment rights to editorial discretion. There is no basis for citing it as a basis for restricting such rights.

*  *  *

The challenged Florida and Texas laws treat social media platforms essentially as company towns. But social media platforms simply do not demonstrate the requisite characteristics sufficient to treat them as company towns whose moderation decisions are subject to court review for viewpoint discrimination. Instead, consistent with their economic function, they are private actors with their own rights to editorial discretion protected from government interference.

III. The Justifications for Common Carriage Regulation Do Not Apply to Social Media Companies.

The law and economics principles described above establish a general rule of the First Amendment that private property owners like social media companies have the right, responsibility, and need in the marketplace to moderate speech on their platforms. It makes no more sense to apply common carriage regulation to social media platforms than it does to treat them as company towns subject to the First Amendment.

Both Florida’s SB 7072 and Texas’s HB 20 are designed to restrict the ability of social media companies to exercise editorial discretion on their platforms. Each State justified its law by comparing social media companies to common carriers. Florida’s legislative findings included the statement that social media platforms should be “treated similarly to common carriers.” Act of May 24, 2021, ch. 2021-32, § 1(6), 2021 Fla. Laws 503, 505. Texas’ legislature found that “social media platforms function as common carriers” and “social media platforms with the largest number of users are common carriers by virtue of their market dominance.” Act of Sept. 9, 2021, ch. 3, § (3)–(4), 2021 Tex. Gen. Laws 3904, 3904.

But simply “[l]abeling” a social media platform “a common carrier .?.?.?has no real First Amendment consequences.” Denver Area Educ. Telecomm. Consortium, Inc. v. FCC, 518 U.S. 727, 825 (1996) (Thomas, J., concurring in the judgment in part and dissenting in part). And nothing about social media platforms justifies the label in any event: Social media platforms do not hold themselves out to the public as common carriers, and social media platforms lack monopoly power.

A. Social media platforms do not hold themselves out to all comers.

Both the Eleventh Circuit in Moody and the Fifth Circuit in Paxton recognized that one characteristic common carriers share is that they hold themselves out as serving all members of the public without individualized bargaining. See Moody, 34 F.4th 1196, 1220 (11th Cir. 2022); Paxton, 49 F.4th at 469.

Major social media companies, however, do not hold themselves out to the public indiscriminately either for users or the type of speech allowed. Unlike a telephone company or the postal service, both of which carry all private communications regardless of the underlying message, social media companies require all users to accept terms of service dealing specifically with speech in order to use the platform. They also maintain the discretion to enforce their rules as they see fit, both curating and editing speech before presenting it to the world.. As the Eleventh Circuit put it in Moody, social media users “are not freely able to transmit messages ‘of their own design and choosing’ because platforms make—and have always made—‘individualized’ content- and viewpoint-based decisions about whether to publish particular messages or users.” Moody, 34 F.4th at 1220 (quoting FCC v. Midwest Video Corp., 440 U.S. 689, 701 (1979)).

Moreover, the very service that online platforms offer to users, and that users accept, is the moderation of speech in one form or another. Instagram allows users to curate feeds of specialized images, and Twitter does the same for specialized microblogs. Without this core moderation service, the services would be essentially useless to users. By contrast, common carriers do not have as a core part of their service the moderation of speech: any moderation of speech is incidental to operation of the service (e.g. removing unruly passengers).

Judge Srinivasan’s concurring opinion in United States Telecom Association v. FCC, 855 F.3d 381 (D.C. Cir. 2017) (denying rehearing en banc), is instructive on this point. The panel there had denied a petition for review of the FCC’s net neutrality order, which applied common carriage regulation to internet service providers. At the rehearing stage, then-Judge Kavanaugh feared the panel’s opinion would allow the government to “impose forced-carriage or equal-access obligations on YouTube and Twitter.” Id. at 433 (Kavanaugh, J., dissenting). Judge Srinivasan sought to allay that fear by explaining: Social media platforms “are not considered common carriers that hold themselves out as affording neutral, indiscriminate access to their platform without any editorial filtering[.]”. Id. at 392 (Srinivasan, J., concurring) (emphasis added). Indeed, even the Internet service providers deemed common carriers there could escape such designation if they acted like social media platforms and exercised editorial discretion and advertised themselves as doing so. See id. at 389-90 (Srinivasan, J., concurring).

Unlike the telegraph, telephone, the postal service, or even email, major social media companies do not hold themselves out to the public as open to all legal speech—they expressly retain their editorial discretion. They have publicly available terms of service that users must agree to before creating profiles that detail what is and is not allowed on their platforms. While common carriers like airlines may be able to eject passengers based upon conduct even where there is a speech element, social media companies retain the right to restrict pure expression that is inconsistent with their community standards. These rules include limitations on otherwise legal speech and disclose that violators may be restricted from use, including expulsion. Br. for Pet’rs, https://netchoice.org/wp-content/uploads/2023/11/No.‌-22-555_NetChoice-and-CCIAs-Brief-Paxton.pdf, at 4-7.

The Fifth Circuit was wrong to minimize social media platforms’ editorial discretion by comparing their efforts to newspapers curating articles and columns. See Paxton, 49 F.4th at 459-60, 492 (noting that more than 99% of content is not reviewed by a human). Miami Herald did not establish a floor on how much a private actor must exercise editorial discretion in order to be protected by the First Amendment. Nor did it specify that a human must review content rather than a company investing in algorithms to help them moderate content. The Fifth Circuit’s reasoning is essentially a “use it or lose it” theory of the First Amendment, which says if social media companies do not aggressively use their editorial discretion rights, then they can lose them. “That is not how constitutional rights work,” however; the “‘use it or lose it’ theory is wholly foreign to the First Amendment.” U.S. Telecom, 855 F.3d at 429 (Kavanaugh, J., dissenting).

Since social media companies do not hold themselves out to the public as open to all speech, they are not common carriers that can somehow be required to carry third party speech contrary to their terms of service.

B. Social media companies lack gatekeeper monopoly power.

Another reason offered for treating social media platforms like common carriers is that some social media companies are alleged to have “dominant market share,” see Biden v. Knight, 141 S. Ct. 1220, 1224 (2021) (Thomas, J., concurring), or in the words of Turner, “gatekeeper” or “bottleneck” market power. See Turner, 512 U.S. at 656.

As shown above, however, Turner is not really about market power but about the unique physical connection that gave cable providers the power to restrict access to content by the flick of a switch. In any case, there is no basis for concluding that social media companies are all monopolists.

A number of major social media companies covered by the Florida and Texas laws are not in any sense holders of substantial market power as measured by share of visits.[9] Neither are companies like reddit, LinkedIn, Tumblr, or Pinterest, who all have even fewer visits. Nonetheless, the challenged laws would apply to such entities based on monthly users at the national level or gross revenue. See Fla. Stat. §501.2041(1)(g)(4) (covered providers must have at least 100 million monthly users or $100 million in gross annual revenue); Tex. Bus. & Com. Code §§ 120.001(1), .002(b) (covered social media platforms have 50 million monthly active users). But raw revenue or user numbers do not show market power. It is, at the very least, market share (i.e., concentration) that could plausibly be instructive—and even then, market power entails a much more complex determination. See, e.g., Brian Albrecht, Competition Increases Concentration, Truth on the Market (Aug. 16, 2023), https://‌truthonthemarket.com/2023/08/16/competition-increases‌-concentration/. As economist Chad Syverson puts it, “concentration is worse than just a noisy barometer of market power. Instead, we cannot even generally know which way the barometer is oriented.” Chad Syverson, Macroeconomics and Market Power: Context, Implications, and Open Questions, 33 J. Econ. Persp. 23, 26 (2019).

Second, there is no legislative finding of market power that would justify either law: just a bare assertion by the Texas legislature that “social media platforms with the largest number of users are common carriers by virtue of their market dominance.” HB 20 § 1(4). That “finding” by the Texas legislature fails to even define a relevant market, let alone establish market shares, or identify any indicia of market power of any players in that market. In then-Judge Kavanaugh’s words, both Florida and Texas failed to “even tr[y] to make a market power showing.” U.S. Telecom, 855 F.3d at 418 (Kavanaugh, J., dissenting); see also FTC v. Facebook, 560 F. Supp. 3d 1, 18 (D.D.C. Jun. 28, 2021) (“[T]he FTC’s bare assertions would be too conclusory to plausibly establish market power”).

The Texas legislature’s bare assertion is considerably weaker than the “unusually detailed statutory findings” the Court relied on in Turner, 512 U.S. at 646,^[10] and is woefully insufficient to permit reliance on this justification for common-carrier-like treatment under the First Amendment.

Conclusion

The First Amendment protects the marketplace of ideas by protecting private ordering of speech rules. For the foregoing reasons, the Court should reverse the decision of the Fifth Circuit in Paxton and affirm the decision of the Eleventh Circuit in Moody.

[1] Amicus curiae affirms that no counsel for any party authored this brief in whole or in part, and that no entity or person other than amici and their counsel made any monetary contribution toward the preparation and submission of this brief.

[2] Throughout this brief, the term “platform” as applied to the property of social media companies is used in the economic sense, as these companies are all what economists call multisided platforms. See David S. Evans, Multisided Platforms, Dynamic Competition, and the Assessment of Market Power for Internet-Based Firms, at 6 (Coase-Sandor Inst. for L. & Econ. Working Paper No. 753, Mar. 2016).

[3] Valtteri Vuorio & Zachary Horne, A Lurking Bias: Representativeness of Users Across Social Media and Its Implications for Sampling Bias In Cognitive Science, PsyArXiv Preprint at 1 (Feb. 2, 2023); see also, e.g., Alessia Antelmi, et al., Characterizing the Behavioral Evolution of Twitter Users and The Truth Behind the 90-9-1 Rule, in WWW ’19: Companion Proceedings of The 2019 World Wide Web Conference 1035 (May 2019).

[4] “For decades, the 18-to-34 age group has been considered especially valuable to advertisers. It’s the biggest cohort, overtaking the baby boomers in 2015, and 18 to 34s are thought to have money to burn on toys and clothes and products, rather than the more staid investments of middle age.” Ryan Kailath, Is 18 to 34 still the most coveted demographic?, Marketplace.com Dec. 8, 2017), https://www.market‌place.org/2017/12/08/coveted-18-34-year-old-demographic.

[5] https://www.cnbc.com/2023/11/17/apple-has-paused-advertising-on-x-after-musk-promoted-antisemitic-tweet.html.

[6] https://www.vanityfair.com/news/2023/08/twitter-advert‌isers-dont-want-nazi-problem.

[7] https://www.nytimes.com/2023/06/05/technology/twitter-ad-sales-musk.html.

[8] https://www.zdnet.com/article/twitter-seeing-record-user-engagement-the-data-tells-a-different-story.

[9] See https://www.statista.com/statistics/265773/market-share-of-the-most-popular-social-media-websites-in-the-us (Facebook at 49.9%, Instagram at 15.85%, X/Twitter at 14.69%, YouTube at 2.29%); https://gs.statcounter.com/social-media-stats/all/‌united-states-of-america (similar numbers).

[10] See also Pub. L. 102-385 § 2(a)(1) (detailing price increases of cable television since rate deregulation, which is inferential evidence of market power); id. § 2(a)(2) (explaining that local franchising regulations and the cost of building out cable networks leave most consumers with only one available option).

RE: Amicus Letter Supporting Review in Liapes v. Facebook, Inc. (No. S282529), From a Decision by the Court of Appeal, First Appellate District, Division 3 (No. A164880)

The International Center for Law & Economics (“ICLE”) is a nonprofit, non-partisan global research and policy center aimed at building the intellectual foundations for sensible, economically grounded policy. ICLE promotes the use of law and economics methodologies and economic learning to inform policy debates and has longstanding expertise evaluating antitrust law and policy. We thank the Court for considering this amicus letter supporting Petitioner Facebook’s petition for review in which we wish to briefly highlight some of the crucial considerations that we believe should be taken into account when looking at the intermediary liability principles that underlie the interpretation of the Unruh Act.

The Court of Appeal’s decision in Liapes v. Facebook has profound implications for online advertising and raises significant legal and practical concerns that could echo beyond the advertising industry. Targeted advertising is a crucial aspect of marketing, enabling advertisers to direct benign, pro-consumer messages to potential customers based on various considerations, including age and gender. The plaintiff’s argument, and the Court of Appeal’s acceptance of it, present a boundless theory of liability, suggesting that any targeted advertising based on protected characteristics is unlawful. This theory of liability, unfortunately, fails to take account of the nature of Facebook as an online intermediary, and the optimal limitations on liability that this requires when weighing the bad acts of third parties against Facebook’s attempt to provide neutral advertising tools to the benefit of millions of users.

The Unruh Act Is Not a Strict Liability Statute

While the Unruh Act prohibits intentional discrimination, California Civil Code §§ 51 and 51.5, California courts have consistently emphasized that the statute does not impose strict liability for all differential treatment. Rather, the Unruh Act allows for distinctions that serve legitimate nondiscriminatory purposes.

Courts have held that the Unruh Act does not bar practices “justified by ‘legitimate business interests.'” Koebke v. Bernardo Heights Country Club, 36 Cal. 4th 824, 851 (2005). The statute prohibits only discrimination that is “arbitrary, invidious or unreasonable.” Javorsky v. Western Athletic Clubs, Inc., 242 Cal. App. 4th 1386, 1395 (2015). Reasonable, nonarbitrary distinctions are therefore permissible. Differential treatment may qualify as reasonable and nonarbitrary if there is a public policy justification for the distinction. For example, discounts for senior citizens have been deemed nonarbitrary because they advance policies like assisting those with limited incomes. Sargoy v. Resolution Trust Corp., 8 Cal. App. 4th 1039, 1044 (1992). And it is “reasonable” discrimination on the basis of age to prevent minors from entering bars and adult bookstores. Koire v. Metro Car Wash, 40 Cal. 3d 24, 31 (1985).

Thus, the Unruh Act does not impose strict liability merely for practices that have a disparate impact. Harris v. Capital Growth Investors XIV, 52 Cal. 3d 1142, 1149 (1991). While the Unruh Act provides robust protections, it was not intended to forbid all differential treatment. Distinctions based on legitimate justifications remain permissible under the statute’s exceptions.

Firms like Meta operate services facilitating billions of interactions between users and advertisers. In this vast, complex environment, interpreting any ad targeting based on protected class membership as a per se Unruh Act violation would amount to imposing de facto strict liability on the online advertising industry. Setting aside the fact that the Unruh Act is not a strict liability statute, drawing the liability line at this point would have drastic practical consequences.

First, a de facto strict liability standard fails to account for the immense scale and complexity of services like Facebook. Given the number of third-party advertisers and users, as well as the speed and quantity of ad auctions, some incidental correlations between ad delivery and protected characteristics are likely inevitable even absent purposeful exclusions. The Court of Appeal’s opinion exposes both advertisers and platforms like Facebook to litigation based on such correlations, on the theory that the correlations may be “probative” of the intentional discrimination the Unruh Act forbids.

Second, advertisers may have many reasonable, nonarbitrary motivations for targeting their ads to certain demographic groups. For example, targeting older people for certain kinds of medicines, or members of religious groups with information about services in their religion. The Court of Appeal’s opinion will lead to extensive, costly litigation about potential justification for such ad targeting, and in the meantime consumers will be deprived of useful ads.

Finally, if any segmentation of ad targets based on protected characteristics triggers Unruh Act violations, online advertising loses an essential tool for connecting people with relevant messages. This impedes commerce without any showing of invidious discrimination.

Although the Unruh Act provides important protections, overbroad interpretations amount to strict liability incompatible with the realities of a massive, complex ad system. Nuance is required to balance anti-discrimination aims with the actual welfare of users of services. In order to properly parse the line between reasonable and unreasonable discrimination when dealing with a neutral advertising service like Facebook and the alleged bad acts of third parties, it is necessary to incorporate the legal principles of intermediary liability into an analysis under the Unruh Act.

Principles of Intermediary Liability

In public policy and legal analysis, a central objective is to align individual incentives with social welfare, thereby deterring harmful behavior and encouraging optimal levels of precaution. See Guido Calabresi, The Cost of Accidents: A Legal and Economic Analysis 26 (1970). In the online context, this principle necessitates a careful examination of intermediary liability, especially for actors indirectly involved in online interactions.

Intermediary liability applies to third parties not directly causing harm but who can influence primary actors’ behavior to reduce harm cost-effectively. This is particularly relevant when direct deterrence is insufficient, and the intermediary can prevent harm more effectively or at a lower cost than direct enforcement. See Reiner Kraakman, Gatekeepers: The Anatomy of a Third-Party Enforcement Strategy, 2 J.L. Econ. & Org. 53, 56-57 (1986). However, not every intermediary in a potentially harmful transaction should be a target for such liability.

The focus is on locating the “least-cost avoider” – the party that can reduce the likelihood of harm at the lowest overall cost. See Harold Demsetz, When Does the Rule of Liability Matter?, 1 J. of Leg. Stud. 13, 28 (1972); see also Kraakman, supra, at 61 (“[t]he general problem remains one of selecting the mix of direct and collateral enforcement measures that minimizes the total costs of misconduct and enforcement”). This approach aims to balance the costs of enforcement against the social gains achieved as well as the losses that flow from the chilling effects of liability.

Imposing liability involves weighing the administrative costs and the potential lost benefits society might enjoy in the absence of liability. See Ronald Coase, The Problem of Social Cost, 3 J.L. & Econ. 1, 27 (1960) (“[W]hat has to be decided is whether the gain from preventing the harm is greater than the loss which would be suffered elsewhere as a result of stopping the action which produces the harm.”). The least-cost avoider is determined by considering whether the reduction in costs from locating liability on that party is outweighed by the losses caused by restricting other activities that flow from that liability. Calabresi, supra at 141.

The internet comprises various intermediaries like interactive computer services, internet service providers, content delivery networks, and advertising networks, which facilitate interactions between users, content platforms, and various service providers. See generally David S. Evans, Platform Economics: Essays on Multi-Sided Businesses (2011). Sometimes, intermediaries are the least-cost avoiders, especially when information costs are low enough for them to monitor and control end users effectively, or when it is difficult or impossible to identify bad actors using those platforms. But this is not always the case.

While liability can induce actors to take efficient precautions, intermediaries often cannot implement narrow precautions due to limited information or control. Facebook’s platform illustrates this challenge: Facebook has limited to no access to information about the motivations or design of every one of the millions of ad campaigns from millions of individual advertisers on its platform at any given time. Thus, avoiding liability risk might entail broad actions like reducing all services, including those supporting beneficial activities. If the collateral costs in lost activity are significant, the benefits of imposing intermediary liability may not justify its implementation.

Here, overbroad liability could end up severely reducing the effectiveness of advertising in general. This could result in 1) less relevant advertisements for users of online services; 2) reduced value to advertising for businesses, harming in particular small businesses which have limited advertising budgets, and 3) less revenue for online services which rely on advertising revenue, pressuring them to increase revenue through other means like higher ad prices and subscriptions.

The individuals and businesses placing advertisements, not the intermediary ad platform, are the primary actors choosing whether and how to use tools for targeting. As we noted above, under the Unruh Act there are permissible uses of targeted advertising, even when focusing on protected classes. The focus in discouraging discrimination should be on primary actors.

It is not hard to locate parties misusing Facebook’s advertising tools in a way that potentially violates the Unruh Act when evidence of discrimination is presented. On the other hand, intermediaries like Facebook will often lack particularized ex ante knowledge of specific discriminatory transactions or direct control over advertisers’ targeting choices. The only avenue for Facebook to comply with broad liability under the Unruh Act is to altogether remove the ability of businesses to use any characteristic that might theoretically trigger Unruh Act liability, which would result in the harms described above. In situations like this, where the intermediary has little ability to effectively police certain misuses of otherwise benign, neutral tools that enhance social welfare, the case for imposing collateral liability is weakened.

Moreover, some statistically disproportionate ad delivery outcomes may be inevitable given the vast scale of platforms like Facebook. Disparate effects should not automatically equate to impermissible discrimination absent purposeful exclusion. The creation of neutral tools for use by advertisers who then use them to break the law does not imply intentionality by Facebook (or any other advertising platform) to break the law. No one would suggest that a hammer company intends for its product to be misused by customers who use it to bludgeon another human being. Nuance is required.

Broad Unruh Act liability risks unintended harms. Imposing a de facto strict liability regime that treats all ad targeting of protected classes as impermissible under the Unruh Act would drive services like Facebook to restrict lawful advertising tools for all users in order to mitigate liability risks. This impairs a large amount of indisputably legal commerce to deter allegedly illegal advertising by a subset of third parties. Moreover, the effects of such a decision would echo not only throughout the advertising ecosystem, but throughout the internet ecosystem in general where intermediaries might provide similar neutral tools that could run afoul of such a broad theory of liability.

Conclusion

The intermediary liability principles outlined above strongly counsel against the overbroad Unruh Act interpretation embraced by the Court of Appeal in the present matter.

The primary actors are the advertisers choosing whether and how to target ads, not Facebook. The Court of Appeal’s broad view wrongly focused on Facebook’s provision of neutral tools rather than advertisers’ specific uses of those tools.

Given the context-dependent nature of an Unruh Act analysis, the Court of Appeal failed to appropriately balance between deterring allegedly illegal acts by advertisers with the potential loss of value from targeted advertising altogether. The proper duties of intermediaries like Facebook should be limited to feasible actions like removing impermissibly exclusionary ads when notified. They should not include disabling essential advertising tools for all users. The Court of Appeal’s overbroad approach would ultimately harm consumer access to targeted advertising.

With the foregoing in mind, we respectfully urge this Court to grant the pending petition for review. Careful examination of the Court of Appeal’s ruling will reveal it strays beyond the Act’s purpose and ignores collateral harms from overdeterrence. Guidance is needed on balancing antidiscrimination aims with the liberty interests of platforms and their users. This case presents an ideal vehicle for this Court to provide that guidance.

Ayear ago, we cautioned that the EU Cybersecurity Certification Scheme for Cloud Services (EUCS) threatened to embed ill-conceived economic protectionism into the EU’s cybersecurity rules. And, indeed, the European Commission, which has made clear its commitment to pursue “digital sovereignty” for the European Union, can claim some preliminary successes on that front.

A recent draft of EUCS shows that the European Union Agency for Cybersecurity (ENISA) heeded the Commission’s call, contrary to ENISA’s own prior recommendations. Most notably, the draft would preclude entities outside the EU and those under foreign ownership or control from receiving  the highest level of cybersecurity certification.

Read the full piece here.

ICLE Director of Innovation Policy Kristian Stout joined fellow panelists Timothy B. Lee and Pamela Samuelson and moderator Brent Skorup to discuss the emerging legal issues surrounding artificial intelligence and its use of works protected under copyright law on a recent episode of the Federalist Society Regulatory Transparency Project’s Fourth Branch Podcast. The full episode is embedded below.

We thank the International Trade Administration (ITA), the National Institute of Standards and Technology (NIST) and the U.S. Patent and Trademark Office (USPTO) for this opportunity to comment on its call for evidence concerning a new framework for standard-essential patents.[1] The International Center for Law and Economics (ICLE) is a nonprofit, nonpartisan research center whose work promotes the use of law & economics methodologies to inform public-policy debates. We believe that intellectually rigorous, data-driven analysis will lead to efficient policy solutions that promote consumer welfare and global economic growth. ICLE’s scholars have written extensively on competition, intellectual property, and consumer-protection policy.

In this comment, we express concerns about global regulatory developments in the standard-essential patent (SEP) industry. The European Union is in the process of considering legislation that would fundamentally alter the landscape of global standards setting, making it more difficult for inventors to enforce their intellectual-property rights.[2] Not only will this legislation have profound ramifications for companies located all over the globe but—as the USPTO’s call for comments recognizes—the EU risks kicking off a global race to the bottom in regulating SEPs that will ultimately harm innovation and slow the diffusion of groundbreaking technologies.

We are concerned that a tit-for-tat response intended to counteract bad policies in the EU (and among other allied nations) is doomed to do more harm than good. Erecting what amount to protectionist barriers—even if in response to similar regulations abroad—would diminish U.S. interests, as well as those of our partners. Instead, the agencies should be seeking opportunities to influence the policy decisions made in foreign jurisdictions, in the hope that those entities will pursue better policies.

For obvious reasons, the way intellectual-property disputes are resolved has tremendous ramifications for firms that operate in standard-reliant industries. Not only do many of the firms in this space derive a large share of their revenue from patents but, perhaps more importantly, the prospect of litigation dictates how firms structure the transfer of intellectual-property assets. In simple terms, ineffectual judicial remedies for IP infringements and uncertainty concerning the resolution of IP disputes discourage firms from concluding license agreements in the first place.

The key role that IP plays in these industries should impel policymakers to proceed with caution. By virtually all available metrics, the current system works. The development of innovative technologies through standards development organizations (SDOs) has led to the emergence of some of the most groundbreaking technologies that consumers use today;[3] and recent empirical evidence suggests that many of the alleged ills that have been associated with the overenforcement of intellectual-property rights simply fail to materialize in industries that rely on standard-essential patents.[4]

At the same time, “there is no empirical evidence of structural and systematic problems of holdup and royalty stacking affecting standard-essential patent (“SEP”) licensing.”[5] Indeed, “[t]he notion that implementers in such innovation–driven industries are being suffocated by an insurmountable patent royalty stack has turned out to be nothing more than horror fiction.”[6] Yet, without a sound basis, the anti-injunctions approach increasingly espoused by policymakers unnecessarily “adds a layer of additional legal complexity and alters bargaining processes, unduly favoring implementers.”[7]

Licensing negotiations involving complex technologies are legally intricate. It is simply not helpful for a regulatory body to impose a particular vision of licensing negotiations if the goal is more innovation and greater ultimate returns to consumers. Instead, where possible, policy should prefer allowing parties to negotiate at arm’s length and to resolve disputes through courts. In addition to maintaining the sometimes-necessary remedy of injunctive relief against bad-faith implementers, this approach allows courts to explore when injunctive relief is appropriate on a case-by-case basis. Thus, over the course of examining actual cases, courts can refine the standards that determine when an injunctive remedy is inappropriate. Indeed, the very exercise of designing ex ante rules and guidelines to inform F/RAND licensing is antagonistic to optimal policymaking, as judges are far better situated and equipped to make the necessary marginal adjustments to the system.

Against this backdrop, our comments highlight several factors that should counsel preserving the rules that currently govern SEP-licensing agreements:

To start, the SEP space is far more complex than many recognize. Critics often assume that collaborative standards development creates significant scope for opportunistic behavior—notably, patent holdup. The tremendous growth of SEP-reliant industries and market participants’ strong preference for this form of technological development, however, suggest these problems are nowhere near as widespread as many believe.

Second, it is important not to overlook the important benefits conferred by existing IP protections. This includes the advantages inherent in pursuing injunctions rather than damages awards.

Third, weakening the protections afforded to SEP holders would also erode the West’s technological leadership over economies that are heavily reliant on manufacturing, and whose policymakers routinely undermine foreign firms’ intellectual-property rights. In short, while IP promotes innovation, weakened patent protection has second-order effects that are often overlooked, such as ceding advantages to China’s manufacturing sector and thereby exacerbating U.S.-China tensions.

Fourth, while mandated transparency in SEP negotiations may appear beneficial, the reality is more complex, as disclosure requirements can have mixed effects. Further, transparency mandates would likely require government interventions, such as essentiality checks, which can be very costly.

Finally, collective SEP rate-setting raises antitrust issues that stem from firms’ need to share sensitive data in order to determine a standard’s value. Vertically integrated SEP holders setting collective royalties on the inputs they manufacture could enable price-fixing and collusion. Safeguards like third-party mediation in patent pools may be needed so that joint SEP rate negotiation does not violate antitrust rules barring competitors from fixing prices.

I.        Regulatory Developments in Foreign Jurisdictions

In their call for comments, the agencies essentially ask whether regulatory developments in foreign jurisdictions threaten U.S. technological leadership in industries that rely on standard-essential patents and, if so, how the United States should respond:

Do the intellectual property rights policies of foreign jurisdictions threaten any of U.S. leadership in international standard setting, U.S. participation in international standard setting, and/or the growth of U.S. SMEs that rely on the ability to readily license standard essential patents?

If responding affirmatively to question 1, what can the Department of Commerce do to mitigate the effects of any adverse foreign policies relating to intellectual property rights and standards? Please clearly identify any such adverse foreign policies with specificity.[8]

Recent regulatory developments in the European Union loom large over the agencies’ two questions. On April 27, the European Commission published its Proposal for a Regulation on Standard Essential Patents (“SEP Regulation”). The SEP Regulation’s proclaimed aims are to ensure that end users—including small businesses and EU consumers—benefit from products based on the latest standardized technologies; make the EU attractive for standards innovation; and encourage both SEP holders and implementers to innovate in the EU, make and sell products in the EU, and be competitive in non-EU markets.[9]

While we share the agencies’ concern, responding to this foreign legislation (and other international responses that are likely to arise) by enacting similar policies would only exacerbate the situation and further erode U.S. technological leadership. In fact, several of the EU legislation’s shortcomings that would be rendered more destructive if the United States responded in kind.

As ICLE-affiliated scholars have explained in comments on the draft European legislation,[10]  the available evidence does not support a finding of market failure in SEP-licensing markets that would justify intrusive regulatory oversight. Instead, the Commission’s own evidence points to the low incidence of SEP litigation and no systemic negative effects on SEP owners and implementers. The mobile-telecommunication market, which is claimed to have the most SEP litigation and licensing inefficiencies, has over the years seen rapid growth, expansion, declining consumer prices, and new market entry.

Some market imperfections are necessary-but-not-sufficient conditions for regulatory intervention. Regulation might not be necessary or proportionate if its aims could be achieved with less costly instruments.

The EU’s proposed SEP Regulation appears to pursue the value-redistributive function of imposing costs on only one group (SEP owners), while accruing all benefits to non-EU (or US)-based standard implementers. It is difficult to find justification for such value redistribution from the evidence presented on the functioning of SEP licensing markets.

The proposed EU SEP Regulation applies to all standards licensed on FRAND terms. It is unclear how many standards would be caught and why all standards licensed on FRAND terms are presumed to be inefficient, requiring regulatory intervention. One early study identified 148 standards licensed on FRAND terms in a 2010 laptop. No evidence was presented that licensing inefficiencies of these standards caused harms in laptop markets.

The EU legislation would require evaluators and conciliators that need to be qualified and experienced experts in relevant fields. There are unlikely to be enough evaluators to conduct essentiality checks reliably on such a massive scale.

To make matters worse, the proposed SEP Regulation raises competition concerns, as it requires SEP owners to agree on global aggregate royalty rates. No safeguards are provided against the exchange of sensitive commercial information and possible cartelization.

There is also a risk that legislation seeking to make the standardization space more transparent, by mandating aggregate royalty-rate notifications and nonbinding expert opinions on global aggregate royalty rates, may lead to even more confusion for implementers.

Finally, the EU’s proposed SEP Regulation would have extraterritorial effects. Indeed, while the SEP Register and system of “essentiality checks” created by the regulation would apply only for patents in force in EU Member States, its system of nonbinding opinions on aggregate royalties and FRAND determination would apply worldwide, covering portfolios in other countries. Other countries—including the United States—may follow suit and introduce their own regulations on SEPs. Such regulations may be used as a strategic and protectionist tool to devaluate the royalties of innovative SEP owners. The proliferation of regulatory regimes would make SEP licensing even more costly, with unknown effects on the viability of the current system of collaborative and open standardization.

Considering the above, it would appear unwise for the United States to mimic the EU’s draft SEP regulation. In its current form, the regulation is likely to harm both U.S. and European innovators. In turn, this threatens the west’s technological leadership on a global stage and will serve the interests of jurisdictions whose economies rely heavily on implementing standard-essential technologies and that generally have weaker IP protection than either the United States or the EU.

Instead, the agencies should look for opportunities to work with their foreign counterparts to improve the proposed EU legislation (and other similar measures in other jurisdictions). Neither EU nor U.S. interests will be well-served by these sorts of regulatory endeavors, least of all if both areas enact ill-advised SEP policies. Sound policy should be focused on ensuring that the successful SEP ecosystem continues to perform as impressively as it has to date. Enacting defensive measures against the EU legislation will create a tit-for-tat dynamic that will double the obstacles faced by innovators in both the EU and the United States, allowing foreign rivals to take advantage of the situation.

II.      Regulatory Restraint

In their call for comments, the agencies ask what private entities can do to boost America’s participation in international standard-setting efforts:

What more can other entities do, such as standards development organizations, industry or consumer associations, academia, or U.S. businesses to help improve American leadership, participation in international standard setting, and/or increased participation of small to medium-sized enterprises that rely on the ability to readily license standard essential patents?[11]

While this is a good way to look at the issue (today’s standardization practices were born of spontaneous market interactions, rather than government fiat, which leaves private entities with a clearly significant role to play in this space), one should not overestimate the extent to which governments can identify inefficiencies that may afflict standard-reliant industries and nudge private parties to resolve them—e.g., by asking SDOs to curb the use of injunctions or encourage collective royalty-setting agreements.

It’s tempting for lawmakers to look at the complex SEP licensing process as a Gordian Knot to be solved through regulatory fiat. But pursuing Alexander’s solution, though expedient, would similarly leave the SEP licensing ecosystem in tatters.

Consider smartphones: Tens of thousands of patents are essential to making smartphone technology work.^[12] Some critics posit that this makes it extremely difficult to market smartphones effectively, but no evidence supports this claim, and the proliferation of smartphones suggests otherwise.[13] It is worth considering that cellphone technology marks the culmination of research efforts spanning the entire globe. The coordinated efforts of these numerous firms are not the result of government coercion, but the free play of competitive forces.

Coordination on such a vast scale is no simple task. And yet, of the vast array of options available to them, an increasing number of firms have settled on one particular paradigm to solve these coordination problems: the development of new technologies and open standards within SDOs. These organizations and their members are responsible not only for wireless cellular technologies (e.g., 3G, 4G, 5G) but also for such high-profile technologies as Wi-Fi, USB, and Blu-ray, among many others.[14]

Throughout history, economic actors have sought to reap the benefits of specialization and interoperability. This has led to the emergence of various standardization practices, ranging from de facto standards and competition for the market, to complex standard-setting procedures within SDOs.

Ultimately, because interoperability standards rely on firms being able to coordinate their behavior, standardization necessarily implies a degree of incentive compatibility. That is, parties will coordinate their behavior only if they expect that doing so will be mutually beneficial. “This mutuality of considerations has been at the heart of the voluntary FRAND bargain from the outset, given that any risks of holdup or misappropriation of information are bilateral—that is, such risks work in both directions.”[15] This implies that SDOs must design balanced internal rules that bring both patent holders and implementers to the table through mutually agreeable interoperability standards, and guarantee that they will continue to work together into the future as new technologies emerge.[16]

Establishment of SDO interoperability standards typically follows a process by which interested parties come together and identify technological problems that they might be able to solve cooperatively.^[17] SDO members include a wide range of stakeholders, including (among others): companies that manufacture products implementing the technology, companies that market services that use the standards, companies that operate networks that practice the standards, technology firms that create technologies that are included in the standards, academic institutions, and government agencies.[18]

The SDO provides information to interested parties about the standard-setting project and a forum for collaboration.[19] Members attend standard-setting meetings, vote on standardization decisions, and make technological contributions. Participation in standard setting can be subject to a substantial fee and always entails considerable time. There are policies and procedures (“bylaws”) that govern the process of adoption and standard development. Participation in SDOs is voluntary and is subject to acceptance of the terms and conditions set out in the bylaws. These aim to allow the most appropriate technology to become standardized, based on several factors. This is a democratic and consensus-based process designed to ensure that no single participant can manipulate it. Many SDOs also allow for post-adoption appeals by dissenting members. This ultimately leads to a series of technical specifications upon which implementers can build products.

Throughout this process, a critical challenge for SDOs is to ensure that their internal regulations remain “incentive compatible.” To optimize their technological output and ensure the success of their standards, SDOs must attract the right mix of both implementers and innovators. “Most succinctly, the ‘right membership’ comprises a significant portion of each class of stakeholder whose active support is needed to achieve broad adoption.”[20] They thus need to design internal procedures that strike a balance between the sometimes-diverging interests of these stakeholders.

This is no simple task. Although there are numerous ways in which these rules may favor a particular group of participants, allocating the profits of standardization is perhaps the most salient. To a first approximation, SEP holders will tend to favor internal rules that allow them to charge prices that are close to the monopoly benchmark (though not the double-marginalization one). Conversely, implementers will generally prefer policies that limit SEP holders’ returns (so long as this does not dry up the supply of inventions). However, these first-order incentives may not always hold true in the real world. Practical considerations may, for instance, urge SEP holders to accept a pricing structure that is not “profit maximizing” in the short run, but which may incentivize further cooperation or the adoption of an underlying technology.^[21]

The above has important consequences for patent and antitrust policy in SEP-reliant industries. As we have explained, collaborative standard development gives rise to complex incentives, as well as a web of heterogeneous and deliberately incomplete contracts (i.e., where the parties choose not to specify some aspects of their agreement).[22] Given this diversity, uniform and centralized policies that needlessly constrain the range of negotiation—such as a federal-enforcement presumption against injunctions—would likely lead to fewer agreements and inefficient outcomes in numerous cases, especially compared to case-by-case adjudication of F/RAND commitments under the common law of contract.[23]

In short, “standards organizations and market participants are better than regulators at balancing the interests of patent holders and implementers.”[24] Interfering with the emergent norms of the standard-development industry thus risks undermining innovators’ expectations of a reasonable return on their investments:

Each of the innovative companies that agrees to be an SSO participant does so with the understanding of the investments they have made in research, development, and participation, as well as the risks that their innovations may not be selected for incorporation in the standard. They bear these investments and risk with the further understanding that they will receive adequate and fair remuneration as part of the FRAND commitment they have made to the SSO.

Unfortunately, the actions of the courts and the proposals by commentators are greatly undermining the value and benefits of SSO participation that are expected….[25]

III.    The Importance of Injunctions

The agencies’ call for comments appears concerned that current standardization practices may be hindering U.S. innovation and the creation of startups in the SEP space:

Are current fair, reasonable, and non-discriminatory (FRAND) licensing practices adequate to sustain U.S. innovation and global competitiveness? Are there other international models which would better serve U.S. innovation in the future?

Are there specific U.S. intellectual property laws or policies that inhibit participation in standards development?

Are there specific U.S. intellectual property laws or policies that inhibit growth of SMEs that rely on licensing and implementing standards? [26]

While they are not mentioned explicitly in the agencies’ call for comments, the role of injunctions sought against implementers by SEP holders looms large over the above questions. The use of injunctions is arguably one of the most contentious—and widely misunderstood—topics in SEP policy debates. While often portrayed as a means for inventors to extract unreasonable royalties from helpless implementors injunctions are, in fact, a critical legal tool that encourages all parties in the standardization space to come to the negotiation table. In fact, even the EU’s draft regulation on SEPs—which in many other respects reduces the protections afforded to inventors—implicitly recognizes the crucial role that injunctions play, by ensuring that the various proposed SEP transparency and arbitration procedures do not undermine parties’ ability to obtain an injunction:

The obligation to initiate FRAND determination should not be detrimental to the effective protection of the parties’ rights. In that respect, the party that commits to comply with the outcome of the FRAND determination while the other party fails to do so should be entitled to initiate proceedings before the competent national court pending the FRAND determination. In addition, either party should be able to request a provisional injunction of a financial nature before the competent court.[27]

A.   The Fundamental Value of Injunctions

Historically, one of the most important features of property rights in general, and patents in particular, is that they provide owners with the power to exclude unauthorized use by third parties and thus enable them to negotiate over the terms on which instances of use or sale will be authorized.^[28] While the ability to exclude is important in creating the incentive to innovate, it is equally—and perhaps more—important in facilitating the licensing of inventions.^[29]

There are many reasons that someone may invent a new product or process. But if they are to be optimally encouraged to distribute that product and thus generate the associated social welfare, it is crucial that they retain the ability to engage supply chains to commercialize the invention fully.^[30] “[T]he patent system encourages and enables not just invention but also innovation by providing the basic, enforceable property rights that facilitate (theoretically) efficient organizations of economic resources and the negotiations necessary to coordinate production among them.”[31] If a patent holder believes that the path to commercialization and remuneration is hindered by infringers, she will have less incentive to invest fully in the commercialization process (or in the innovation in the first place).

Removing the injunction option… not only changes the bargaining range (and makes infringement a valid business option), but, by extension, it lowers the expected returns of investing in the creation and commercialization of patents, in the first place…. With a no-injunction presumption…, as long as the expected cost of litigation is less than the expected gain from infringing without paying any royalties, potential licensees will always have an incentive to pursue this strategy. The net result is a shift in bargaining power so that, even when license agreements are struck, royalty rates are lower than they would otherwise be, as well as an increased likelihood of infringement.[32]

Because infringement affects both the initial incentive to innovate as well as the complex process of commercialization, courts have historically granted injunctions against those who have used a patent without proper authorization.

B.   Damages Alone Are Often Insufficient

Injunctions are almost certainly the most powerful means to enforce property rights and remedy breaches. Nonetheless, courts may sometimes award damages, either in addition to or as an alternative to awarding an injunction.^[33] It is often difficult to establish the appropriate size of an award of damages, however, when intangible property—such as invention and innovation, in the case of patents—is the core property being protected.

In this respect, a key feature of patents is that they possess uncertain value ex ante. The value of a particular invention or discovery cannot be known until it is either integrated into the end-product that will be distributed to consumers, or actually used by consumers.[34] This massive upfront uncertainty creates the need for technology designers to carefully structure their investments such that the risk/reward ratio remains sufficiently low. This, in turn, means ensuring that their inventions’ commercialization can reasonably be expected to generate sufficient profits.

Commercializing highly complex innovations, such as pharmaceuticals and advanced technologies, requires a large degree of risk taking and capital investment, as well as massive foregone opportunities. As such, it will often be difficult, or even impossible, to adequately calculate appropriate monetary damages for the unauthorized use of a patent, even if the patent’s ex post value is knowable. Put differently, the inability to bargain effectively for royalties post-standardization may “deter investment… and ultimately harm consumers.”[35]

While it is necessary to establish damages for violations after the fact, it will nearly always be appropriate to award injunctions to deter ongoing violations. This would further allow the property owner to do their own value calculations, based on their investments, sunk costs, and—critically—lost opportunities that were foregone in order to realize the particular invention. “[A] property rule is superior to a liability rule when ‘the court lacks information about both damages and benefits.’ Without accurate information, the damages may be set below the actual level of harm, encouraging the ‘injurer’ (or infringer) to engage in an excessive level of activity—in our case, increased infringement.”[36]

C.   Injunctions Encourage Efficient Licensing Negotiations

In addition to the concerns outlined in the previous section, it is worth noting that curbs on injunctions pertaining to SEPs would make inventors bear the risk of opportunistic behavior, thus enabling  firms to opt out of commercial negotiations and wait for potential litigation. In turn, this would tilt the bargaining scale in their favor in subsequent royalty negotiations undertaken in the shadow of prior court proceedings.[37]

The U.S. Supreme Court’s 2006 decision in eBay v. MercExchange offers a case in point. The court rejected the “general rule” that a prevailing patentee is entitled to an injunction.[38] In the aftermath of the decision, courts refused to grant injunctions in considerably more cases.[39]

Nearly two decades later, however, questions remain regarding eBay’s effect on patent licensing, negotiation, and litigation.[40] In particular, it is likely that eBay systematically distorted the relative bargaining positions in SEP licensing in favor of implementers, at the expense of patent holders. One post-eBay assessment argues that limiting injunctions to prevent holdup results in more “false positives”—where patent holders with no designs of patent holdup are nonetheless denied injunctive relief—than it does deterrence of actual holdups.[41] The result is a reduction in the cost of willful infringement and “under-compensation” for innovation.[42]

One of the important features of injunctions that critics miss is that they are not solely a tool for simple exclusion from property, but a tool that promotes efficient bargaining.[43] If a property holder ultimately has the right to exclude infringers, there is relatively more weight placed on the importance of initial bargaining for licenses. “It is the very threat of the injunction right—and its associated high transaction costs—that brings the parties to the negotiating table and motivates them to draw upon the full scope of their knowledge and creativity in forming contractual and institutional solutions to the perceived holdup problem.”[44]

Post-eBay, “efficient infringement” becomes a viable choice for firms seeking to maximize profits. Thus, implementing firms seeking to pay as little as possible for use of an invention have incentive to disregard the bargaining process with a patent holder altogether. The relative decline in the importance of injunctions narrows the bargaining range. The narrower range of prices that an implementing firm will offer means that, even where it does bargain, agreement will be less likely. Where rightsholders can be reasonably expected to enforce their patent rights, by contrast, the bargaining range is expanded and agreement is more likely, because the initial cost of negotiating for a license is relatively less than always (or usually) opting for “efficient infringement”; that is, infringement becomes less efficient.

The ultimate tension is not between seeking damages or an injunction, but between whether a firm opts to negotiate or to litigate, while facing the risk of some combination of damages and injunction on the back end.

This reality is particularly important in the context of SDOs, where implementers and innovators are in a constant dance both to maximize their own profits as well as to facilitate the product of an incomplete, joint agreement that binds each party. “The seminal example of intentional contractual incompleteness is the F/RAND commitment common in many [SDO’s] IPR policies.”[45] Permitting one party, through weakened legal doctrine, to circumvent or artificially constrain the bargaining process inappropriately imbalances the careful commercial relationships that should otherwise exist.

In the SEP context, furthermore, it is rarely mentioned that “an implementer’s decision to reject a certifiably F/RAND license and continue to infringe is contrary to the spirit of the F/RAND framework as well.”^[46]

Moreover, it is not typically the case that a negotiation process would end with an injunction and a refusal to license, as critics sometimes allege. Rather, the threat of an injunction is important in hastening an infringing implementer to the table and ensuring that protracted litigation to determine the appropriate royalty (which is how such disputes do actually end) is costly not only to the patentee, but also to the infringer. As James Ratliff and Daniel Rubinfeld explain:

[T]he existence of that threat does not lead to holdup as feared by those who propose that a RAND pledge implies (or should embody) a waiver of seeking injunctive relief. If RAND terms are reached by negotiation, the negotiation is not conducted in the shadow of an injunctive threat but rather in the shadow of knowledge that the court will impose a set of terms if the parties do not reach agreement themselves. The crucial element of this model that substantially diminishes the likelihood that the injunctive threat will have real bite against an implementer willing to license on RAND terms is the assumption that an SEP owner maintains its obligation to offer a RAND license even if its initial offer is challenged by the implementer and, further, even if the court agrees with the SEP owner that its initial offer was indeed RAND. Thus any implementer that is willing to license on court-certified RAND terms can avoid an injunction by accepting those RAND terms without eschewing any of its challenges to the RAND-ness of the SEP owner’s earlier offers.[47]

Ultimately, this means that an implementer that accepts nominally F/RAND terms need not be an actual “willing licensee,” but instead can gain that designation as a matter of law without ever accepting a royalty rate within the true bargaining range that includes the licensor’s valid injunction threat. “[B]y stripping the SEP holder’s right to injunctive relief, [a no-injunction rule] may enable a potential licensee to delay good faith negotiation of a F/RAND license and the patent holder could be forced to accept less than fair market value for the use of the patent…. Undermining this bargaining outcome using antitrust rules runs a significant risk of doing more harm than good.”[48]

For the purposes of this proceeding, the lesson is clear: U.S. policy needs to return to a neutral position in which both parties in a F/RAND negotiation are encouraged to reach mutually agreeable terms in arm’s length licensing transactions. The effects of eBay and its progeny have distorted that bargaining process. Here, the agencies have an important role to play in pressing the need for this change.

IV.    Increased Transparency Is No Free Lunch

The agencies’ call for comments asks whether increased transparency requirements in the SEP space could make SEP licensing more efficient:

What can the Department of Commerce do to mitigate emergence or facilitate the resolution of FRAND licensing disputes? Can requiring further transparency concerning patent ownership make standard essential patent (SEP) licensing more efficient? What are other impediments to reaching a FRAND license that the Department of Commerce could address through policy or regulation?[49]

But while fostering transparency may appear to be a win-win proposition for all parties in the standardization space, the reality is far more complex. In many instances, inventors are already required to disclose their standard essential patents—and these requirements have ambiguous effects.[50] Given this, demands for further transparency would almost certainly entail some form of government intervention, such as the creation of SEP registers and government-run essentiality checks, which seek to verify whether the patents that firms declare as standard-essential are truly so.

Unfortunately, these attempts to make SEP-reliant industries more transparent are anything but costless. The EU’s draft SEP regulation offers a case in point. The regulation would create a system of government-run essentiality checks and nonbinding royalty arbitrations that seek to make the process easier for implementers. But as ICLE scholars have written, this scheme will prove extremely difficult and costly to operate in practice.[51] Much the same would be true of attempts to introduce similar measures in the United States.

The proposed EU regulation would rely on qualified experts to work as evaluators and conciliators. Evaluators will need specialized knowledge of the particular technological area in which they will conduct essentiality checks. The European Commission estimates that there are about 1,500 experts (650 patent attorneys and 800 patent examiners) qualified to do essentiality checks in the EU.[52]

The sheer magnitude of the task, however, will require many more evaluators and it is very doubtful that the optimal number of potential qualified experts are even available to join this process. For certain, special arrangements would need to be made with patent offices to grant patent examiners leave to conduct essentiality checks. Each year, evaluators would need to test a random sample of up to 100 SEPs if requested by each SEP owner or an implementer per standard. Thus, the amount of work may exponentially increase depending on how many standards are caught by the regulation.

If 148 FRAND-licensed standards per laptop are to serve as a rough proxy, then we might expect more than 100-200 standards to be checked for essentiality every year. In addition, if SEP owners and implementers regularly use the possibility of testing up to 100 SEPs per standard and per SEP owner, the sheer magnitude of work may exceed the capacity of patent attorneys. Patent attorneys may find it challenging to regularly engage in such high volumes of essentiality checks while also serving other clients. And why should they do it at all unless the rate of pay is at least what they could earn in a patent law firm? To be blunt, the work would not be as much fun as acting for real clients, so the pay would probably have to be even higher to attract applicants.

Consequently, it is very unlikely that the capability even exists to annually perform a large number of essentiality checks of registered SEPs. If the requirements to become an evaluator were relaxed to address this workload, this would cast doubt on the reliability of the whole system. There is no point in building a battleship unless you are sure you can get a competent crew.

Additionally, the patent attorneys most likely to be familiar with these technologies may well also find themselves with conflicts of interest. They will probably have worked for some SEP owners or implementers. Elaborate rules to avoid such conflicts would need to be implemented to prevent patent attorneys who were, or still are, engaged with certain clients from becoming evaluators of those clients’ registered SEPs. The conflicts problem would, of course, apply not just to individual attorneys but to their entire firms.

Conciliators would also need to be experts in the field. They might come from the ranks of retired judges, seasoned former company officials, or experienced lawyers. Conflict-of-interest provisions would also be needed to ensure their independence and impartiality in mandatory FRAND determinations.  But the job would, again, have to be sufficiently attractive, both in remuneration and in work content and culture. The Commission has made no investigation as to whether a sufficiently large pool of credible individuals could be found to make the system work.

Of course, there are well-established voluntary systems of conciliators and mediators, some of which are used now to help resolve FRAND disputes. But the proposal adds the idea of compulsory mediation or conciliation. There is scant evidence that either system works in other commercial disputes around the world, and it is unclear why it should be assumed to work here.

In short, it is doubtful that a government-operated scheme of essentiality checks and SEP-royalty arbitrations could reach satisfactory outcomes, as the expertise to do so is lacking and attracting potentially thousands of professionals from the private sector would be too costly. The result is that any government scheme along these lines is unlikely to have the necessary staffing to conduct its mission to the requisite standard. It would thus risk doing more harm than good.

V.      SEP Rights and China’s ‘Cyber Great Power’ Ambitions

In their call for comments, the agencies express a desire to protect the United States’ leading position in the field of standard development and implementation:

Are there steps that the Department of Commerce can take regarding intellectual property rights policy that will help advance U.S. leadership in standards development and implementation for critical and emerging technologies?[53]

The agencies essentially ask what active steps they could take to preserve the United States’ leading position. This, however, ignores the arguably more important question: What steps should the United States avoid taking? As we explain below, U.S. agencies should be particularly careful not to weaken intellectual-property protection in ways that may, ultimately, favor firms in other jurisdictions, such as China.

Observers often regard intellectual property as merely protecting original creations and inventions, thus boosting investments. But while IP certainly does this, it is important to look beyond this narrow framing. Indeed, by protecting these creations, intellectual-property protection—particularly that of patents—produces beneficial second-order effects in several important policy areas.

Consequently, weakening patent protection could have detrimental ramifications that are routinely overlooked by policymakers. This includes giving a leg up to jurisdictions that are heavily geared toward manufacturing, rather than  R&D, and specifically to  China (with knock-on effects for ongoing political tensions between these two superpowers).

As the USPTO has observed:

Innovation and creative endeavors are indispensable elements that drive economic growth and sustain the competitive edge of the U.S. economy. The last century recorded unprecedented improvements in the health, economic well-being, and overall quality of life for the entire U.S. population. As the world leader in innovation, U.S. companies have relied on intellectual property (IP) as one of the leading tools with which such advances were promoted and realized.[54]

The United States is a world leader in the production and commercialization of IP, and naturally seeks to retain that comparative advantage.[55] IP and its legal protections will become increasingly important if the United States is to maintain its prominence, especially when dealing with international jurisdictions, like China, that don’t offer similar levels of legal protection.[56] By making it harder for patent holders to obtain injunctions, licensees and implementers gain the advantage in the short term, because they are able to use patented technology without having to engage in negotiations to pay the full market price. In the case of many SEPs—particularly those in the telecommunications sector—a great many patent holders are U.S.-based, while the lion’s share of implementers are Chinese. Potential anti-injunction policies may thus amount to a subsidy to Chinese infringers of western technology.

At the same time, China routinely undermines western intellectual-property protections through its industrial policy. The government’s stated goal is to promote “fair and reasonable” international rules, but it is clear that China stretches its power over intellectual property around the world by granting “anti-suit injunctions” on behalf of Chinese smartphone makers, designed to curtail enforcement of foreign companies’ patent rights.[57]

In several recent cases, Chinese courts have claimed jurisdiction over F/RAND issues.[58] In Oppo v. Sharp, the Supreme People’s Court of China determined that Chinese courts can set the global terms of what is a fair and reasonable price for a license,[59] even if that award would be considerably lower than in other jurisdictions. This decision follows Huawei v. Conversant, in which a Chinese court for the first time claimed the ability to issue an anti-suit injunction against the Chinese company.[60]

All of this is part of the Chinese government’s larger approach to industrial policy, which seeks to expand Chinese power in international trade negotiations and in global standards bodies.[61] As one Chinese Communist Party official put it: “Standards are the commanding heights, the right to speak, and the right to control. Therefore, the one who obtains the standards gains the world.”[62] Chinese President Xi Jinping frequently (but only domestically) references China’s “cyber great power” ambitions: “We must accelerate the promotion of China’s international discourse power and rule-making power in cyberspace and make unremitting efforts towards the goal of building a cyber great power.”[63] Chinese leaders are intentionally pursuing a two-track strategy of taking over standards bodies and focusing on building platforms to create path dependencies that cause others to rely on Chinese technology.[64] As a Hinrich Foundation Report notes:

Trade and technical standards are inherently interrelated. They are mutually reinforcing. But Beijing treats standard setting, and standards organizations, as competitive domains. This approach risks distorting global trade. Beijing does not support a neutral architecture where iterative negotiating strives for technical interoperability. Instead, Beijing promotes an architecture that bolsters and cements Chinese competitiveness. Due to China’s size and centralization, the consequences of this approach will reverberate across the international system. Given the nature of emerging technology and standards, the consequences will endure.[65]

Insufficient protections for intellectual property will hasten China’s objective of dominating collaborative standard development in the medium- to long-term.[66] U.S. entrepreneurs are able to engage in the types of research and development that drive innovation because they can monetize those innovations. Reducing the returns for patents that eventually become standards will lead to less investment in those technologies. It will also harm the competitive position of American companies that refrain from collaborating because the benefits don’t outweigh the costs, including “missing the opportunity to steer a standard in the manner most compatible with a company’s product offerings, falling behind competitors, or failing to head off broad adoption of a second standard….”[67]

Simultaneously, this will engender a switch to greater reliance on proprietary, closed standards rather than collaborative, open standards. Proprietary standards (and competition among those standards) are sometimes the most efficient outcome: for instance, when the costs of interoperability outweigh the benefits. The same cannot be said, however, for government policies that effectively coerce firms into adopting proprietary standards by raising the relative costs of the collaborative standard-development process. In other words, there are social costs when firms are artificially prevented from taking part in collaborative standard setting and forced instead to opt for proprietary standards.

Yet this is precisely what will happen to U.S. firms if IP rights are not sufficiently enforceable. Indeed, as explained above, collaborative standardization is an important driver of growth.[68] It is crucial that governments do not needlessly undermine these benefits by preventing American firms from competing effectively in these international markets.

These harmful consequences are magnified in the context of the global technology landscape, and in light of China’s strategic effort to shape international technology standards.[69] With U.S. firms systematically deterred from participating in the development of open technology standards, Chinese companies, directed by their government authorities, will gain significant control of the technologies that will underpin tomorrow’s digital goods and services. The consequences are potentially catastrophic:

The effect of [China’s] approach goes far beyond competitive commercial advantage. The export of Chinese surveillance and censorship technology provides authoritarian governments with new tools of repression. Governments that seek to control their citizens’ access to the internet are supportive of Beijing’s “cyber sovereignty” paradigm, which can lead to a balkanized internet riddled with incompatibilities that impede international commerce and slow technological innovation. And when cyber sovereignty is paired with Beijing’s push to redefine human rights as the “collective” rights of society as defined by the state, authoritarian governments gain a shield of impunity for violations of universal norms.[70]

With Chinese authorities joining standardization bodies and increasingly claiming jurisdiction over F/RAND disputes, there should be careful reevaluation of the ways weakened IP protection would further hamper the U.S. position as  a leader in intellectual property and innovation.

To return to the framing question, yes, there are steps the agencies could take to secure and promote U.S. leadership in intellectual-property-intensive industries. The first step, as noted throughout this comment, is to refrain from promoting policies that unnecessarily imbalance the negotiation process between innovators and implementers. The second step is twofold. First, work with trustworthy partners, like the EU, to make sure that U.S. Allies’ IP policies are in alignment with and are geared  toward promoting neutral standards that allow tech industries to thrive. The second part is to advocate for trade policies that dissuade countries like China from using their domestic courts and regulatory agencies as protectionist entities designed simply to advance China’s national interests.

VI.    Competition Concerns with Aggregate Royalties

In the call for comments, the agencies ask:

Do policy solutions that would require SEP holders to agree collectively on rates or have parties rely on joint negotiation to reach FRAND license agreements with SEP holders create legal risks? Are there other concerns with these solutions?[71]

A host of competition concerns are implicated in this question, in that it requires SEP owners to negotiate and ultimately agree on aggregate royalty rates for standards.  This may require SEP owners to exchange sensitive commercial information relevant to establishing the value that devices derive from using the standardized technology. Competition-sensitive data could include projected revenues on a per-unit basis following the incorporation of connectivity in the end products, the number of end products sold on the market, actual and forecast sales, and price projections.[72] The competitive dangers inherent in this process are more serious for those vertically integrated SEP owners, who simultaneously hold SEPs and manufacture standard-implementing products. They would effectively agree to set the costs (royalties) for their inputs and exchange data about their downstream sales.

Jointly negotiated rates could therefore potentially run afoul of antitrust laws that prohibit companies from engaging in price-fixing and collusion. Requirements to jointly negotiate aggregate royalty rates should thus be accompanied by safeguards and guidance that ensure such negotiations comply with antitrust law. An example would be royalty-rate negotiations in patent pools, where pool administrators take a mediatory role, collecting and protecting confidential information from pool members.[73]

It is also unclear whether these joint royalty negotiations would be of much use to either inventors or implementers. For example, the EU has proposed introducing an aggregate notification regulation along these lines. The regulation appears to allow multiple groups of SEP owners to jointly notify their views concerning the appropriate royalties for a given technology. This could add even more confusion for standard implementers. For example, some SEP owners could announce an aggregate rate of $10 per product, another 5% of the end-product price, while a third group would prefer a lower $1 per-product rate.

Moreover, it is unclear how joint aggregate royalty-rate notifications would change the existing practice of unilateral announcement of licensing terms. Many SEP owners already publicly announce their royalty programs in advance. To be on the safe side, SEP owners may simply notify their maximum preference, knowing that negotiations would lead to different prices depending on the unique details of various licensees. As a result, aggregate royalty rates may not produce meaningful data points.

Nonbinding expert opinions on global aggregate royalty rates could also add to the confusion. Implementers would likely initiate the process, which would then proceed in parallel with SEP owners’ joint notifications of aggregate rates. All these differing and possibly conflicting estimates might lead to even greater uncertainty. Moreover, if those providing nonbinding opinions are not universally regarded as “experts,” the parties are unlikely to respect such opinions.

Aggregate royalty notifications and nonbinding opinions might be used in the top-down method for FRAND-royalty determinations. A top-down method provides that the SEP owner should receive a proportional share of a standard’s total aggregate royalty. It requires establishing a cumulative royalty for a standard and then calculating the share of the total royalty for an individual SEP owner. This may be the reason for having aggregate royalty-rate notifications and opinions. At the same time, essentiality checks would still be needed to filter out which patents are truly essential, and to assess each individual SEP owner’s share.

We caution strongly against relying too heavily on the top-down approach for FRAND-royalty determinations. It is not used in commercial-licensing negotiations, and courts have frequently rejected its application. Industry practice is to use comparable licensing agreements. The top-down approach was applied in Unwired Planet v Huawei only as a cross-check for the rates derived from comparable agreements.[74] TCL v Ericsson relied on this method, but was vacated on appeal.[75] The most recent Interdigital v Lenovo judgment considered and rejected its use, finding “no value in Interdigital’s Top-Down cross-check in any of its guises.”[76] Moreover, the top-down approach, as currently applied, relies solely on patent counting. It fails to consider that not every patent is of equal value, nor that some patents may be invalid or not infringed by a specific device.

In short, there are important legal and practical obstacles to the joint negotiation of aggregate royalty rates. Legal mandates to conduct such negotiations would thus be of dubious added value to players in standard-reliant industries.

 

 

 

 

[1] U.S. Patent and Trademark Office, Joint ITA-NIST-USPTO Collaboration Initiative Regarding Standards, Federal Register (Sep. 27, 2023), https://www.federalregister.gov/documents/2023/09/27/2023-20919/joint-ita-nist-uspto-collaboration-initiative-regarding-standards; U.S. Patent and Trademark Office, Joint ITA–NIST–USPTO Collaboration Initiative Regarding Standards; Notice of Public Listening Session and Request for Comments, Federal Register (Sep. 11, 2023), available at https://www.govinfo.gov/content/pkg/FR-2023-09-11/pdf/2023-19667.pdf (“Call for Comments”).

[2] European Commission, Explanatory Memorandum for Proposal for a Regulation of the European Parliament and of the Council on Standard Essential Patents and Amending Regulation (EU) 2017/1001, COM (2023) 232 Final (“Explanatory Memorandum”).

[3] See, e.g., Dirk Auer & Julian Morris, Governing the Patent Commons, 38 Cardozo Arts & Ent. L.J. 294 (2020).

[4] See, e.g., Alexander Galetovic, Stephen Haber & Ross Levine, An Empirical Examination of Patent Holdup, 11 J. Competition L. & Econ. 549 (2015). This is in keeping with general observations about the dynamic nature of intellectual property protections. See, e.g., Ronald A. Cass & Keith N. Hylton, Laws of Creation: Property Rights in the World of Ideas 42-44 (2013).

[5] Oscar Borgogno & Giuseppe Colangelo, Disentangling the FRAND Conundrum, DEEP-IN Research Paper (Dec. 5, 2019) at 5, available at https://ssrn.com/abstract=3498995.

[6] Richard A. Epstein & Kayvan B. Noroozi, Why Incentives for “Patent Holdout” Threaten to Dismantle FRAND, and Why It Matters, 32 Berkeley Tech. L.J. 1381, 1411 (2017).

[7] Borgogno & Colangelo, supra note 5, at 5.

[8] Call for Comments, supra note 1, Questions 1 and 2.

[9] Proposal for a Regulation of the European Parliament and of the Council on Standard Essential Patents and Amending Regulation (EU) 2017/1001, COM (2023) 232 Final (“Draft SEP Regulation”).

 

[10] Robin Jacob & Igor Nikolic, ICLE Comments Regarding the Draft Regulation on Standard Essential Patents (Jul. 28, 2023), available at https://laweconcenter.org/wp-content/uploads/2023/07/ICLE-Comments-to-the-SEP-Regulation.pdf.

[11] Call for Comments, supra note 1, Question 3.

[12] See, e.g., Jorge Padilla, John Davies, & Aleksandra Boutin, Economic Impact of Technology Standards: The Past and the Road Ahead (2017), available at https://www.compasslexecon.com/wp-content/uploads/2018/04/CL_Economic_Impact_of_Technology_Standards_Report_FINAL.pdf (Section 3 has an in-depth discussion of the adoption of standards and the benefits to the growth of mobile technology); iRunway, Patent & Landscape Analysis of 4G-LTE Technology 9-12 (2012), https://www.i-runway.com/images/pdf/iRunway%20-%20Patent%20&%20Landscape%20Analysis%20of%204G-LTE.pdf.

[13] See, e.g., Galetovic, et al., supra note 4; Keith Mallinson, Don’t Fix What Isn’t Broken: The Extraordinary Record of Innovation and Success in the Cellular Industry under Existing Licensing Practices, 23 Geo. Mason L. Rev. 967 (2016); Damien Geradin, Anne Layne-Farrar, & Jorge Padilla, The Complements Problem within Standard Setting: Assessing the Evidence on Royalty Stacking, 14 B.U. J. Sci. & Tech. L.144 (2008).

[14] Auer & Morris, supra note 3, at 5.

[15] Epstein & Noroozi, supra note 6, at 1394.

[16] See, e.g., Daniel F. Spulber, Standard Setting Organisations and Standard Essential Patents: Voting and Markets, 129 Econ. J. 1477, 1502-03 (2018) (“The interaction between inventors and adopters helps explain the variation of decision rules among SSOs, ranging from majority rule to consensus requirements…. Technology standards will be efficient when SSO decision making reflects the countervailing effects of voting power and market power.”).

[17] See Kirti Gupta, How SSOs Work: Unpacking the Mobile Industry’s 3GPP Standards, in The Cambridge Handbook of Technical Standardization Law: Competition, Antitrust, and Patents (Jorge L. Contreras ed., 2017).

[18] See Kristen Jakobsen Osenga, Ignorance Over Innovation: Why Misunderstanding Standard Setting Organizations Will Hinder Technological Progress, 56 U. Louisville L. Rev. 159, 178 (2018); Andrew Updegrove, Value Propositions, Roles and Strategies: Participating in a SSO, in The Essential Guide to Standards, https://www.consortiuminfo.org/guide (last visited Jan. 23, 2022).

[19] Adapted from Auer & Morris, supra note 3, at 18-19.

[20] Andrew Updegrove, Business Considerations: Forming and Managing a SSO, in The Essential Guide to Standards, https://www.consortiuminfo.org/guide/forming-managing-a-sso/business-considerations (last visited Nov. 6, 2023).

[21] See, e.g., Jonathan M. Barnett, The Host’s Dilemma: Strategic Forfeiture in Platform Markets for Informational Goods, 124 Harv. L. Rev. 1861, 1883 (2010) (showing that firms routinely forfeit their intellectual assets in order to boost the growth of the platform they operate).

[22] See Joanna Tsai & Joshua D. Wright, Standard Setting, Intellectual Property Rights, and the Role of Antitrust in Regulating Incomplete Contracts, 80 Antitrust L.J. 157, 159 (2015) (“SSOs [standard-setting organizations] and their IPR policies appear to be responsive to changes in perceived patent holdup risks and other factors. We find the SSOs’ responses to these changes are varied, and that contractual incompleteness and ambiguity persist across SSOs and over time, despite many revisions and improvements to IPR policies. We interpret the evidence as consistent with a competitive contracting process and with the view that contractual incompleteness is an intended and efficient feature of SSO contracts.”) (emphasis added).

[23] See, e.g., Daniel F. Spulber, Licensing Standard Essential Patents with FRAND Commitments: Preparing for 5G Mobile Telecommunications, 18 Co. Tech. L.J. 79, 147 (2020) (“Adjudication of SEP disputes guided by common law principles and comparable licenses complements SSO FRAND commitments and market negotiation of SEP licenses. Adjudication based on common law and comparable licenses provides general rules for the resolution of SEP disputes that does not restrict SSO IP policies and or interfere with consensus decision making by SSOs. Such adjudication also does not interfere with efficient market negotiation of SEP licenses.”).

[24] Id. at 148.

[25] Osenga, supra note 19, at 213-14.

[26] Call for Comments, supra note 1, Questions 4, 5, 6.

[27] Draft SEP Regulation, preamble at (35).

[28] Richard A. Epstein, The Clear View of The Cathedral: The Dominance of Property Rules, 106 Yale L.J. 2091, 2091 (1996) (“Property rights are, in this sense, made absolute because the ownership of some asset confers sole and exclusive power on a given individual to determine whether to retain or part with an asset on whatever terms he sees fit.”)

[29] See generally Edmund W. Kitch, The Nature and Function of the Patent System, 20 J.L. & Econ. 265 (1977); F. Scott Kieff, Property Rights and Property Rules for Commercializing Inventions, 85 Minn. L. Rev. 697 (2001).

[30] See, e.g., Barnett, supra note 22, at 856 (“Strong patents provide firms with opportunities to disaggregate supply chains through contract-based relationships, which in turn give rise to trading markets in intellectual resources, whereas weak patents foreclose those options.”).

[31] Dirk Auer, Geoffrey A. Manne, Julian Morris, & Kristian Stout, The Deterioration of Appropriate Remedies in Patent Disputes, 21 Federalist Soc’y Rev. 158, 160 (2020).

[32] Id. at 163.

[33] See, e.g., Doris Johnson Hines & J. Preston Long, The Continuing (R)evolution of Injunctive Relief in the District Courts and the International Trade Commission, IP Litigator (Jan./Feb. 2013) (citing Tracy Lee Sloan, The 1988 Trade Act and Intellectual Property Cases Before the International Trade Commission, 30 Santa Clara L. Rev. 293, 302 (1990) (“Out of 221 intellectual property cases between 1974 and 1987, the ITC found that only five failed to establish sufficient injury… for injunctive-type relief.”)), available at https://www.finnegan.com/en/insights/articles/the-continuing-r-evolution-of-injunctive-relief-in-the-district.html.

[34] And even then, the specific contribution of a particular patent to ultimate consumer value will remain uncertain. See Robert P. Merges, Of Property Rules, Coase, and Intellectual Property, 94 Colum. L. Rev. 2655, 2659 (1994) (“The problems with [clearly defining harms/benefits] in the IPR field result from the abstract quality of the benefits conferred by prior works and the cumulative, interdependent nature of works covered by IPRs. Valuation, then, is at least as great a problem as detection.”)

[35] See Richard Epstein, F. Scott Kieff, & Daniel Spulber, The FTC, IP, and SSOs: Government Hold-Up Replacing Private Coordination, 8 J. Competition L. & Econ. 1 (2012) at 21, available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1907450 (“The simple reality is that before a standard is set, it just is not clear whether a patent might become more or less valuable. Some upward pressure on value may be created later to the extent that the patent is important to a standard that is important to the market. In addition, some downward pressure may be caused by a later RAND commitment or some other factor, such as repeat play. The FTC seems to want to give manufacturers all of the benefits of both of these dynamic effects by in effect giving the manufacturer the free option of picking different focal points for elements of the damages calculations. The patentee is forced to surrender all of the benefit of the upward pressure while the manufacturer is allowed to get all of the benefit of the downward pressure.”).

[36] Merges, supra note 38, at 2666-67 (quoting A. Mitchell Polinsky, Resolving Nuisance Disputes: The Simple Economics of Injunctive and Damage Remedies, 32 Stan. L. Rev. 1075, 1092 (1980)).

[37] See Auer, et al., supra note 35, at 163 (“It also establishes this lower royalty rate as the ‘customary’ rate, which ensures that subsequent royalty negotiations, particularly in the standard-setting context, are artificially constrained.”).

[38] eBay Inc. v. MercExchange, LLC, 547 U.S. 388 (2006).

[39] See Benjamin Petersen, Injunctive Relief in the Post-eBay World, 23 Berkeley Tech. L.J. 193, 196 (2008), (“In the two years after the Supreme Court’s ruling in eBay, there were thirty-three district court decisions that interpreted eBay when determining whether to grant injunctive relief to a patent holder. Of these decisions, twenty-four have granted permanent injunctions and ten have denied injunctions.”). See also Bernard H. Chao, After eBay, Inc. v. MercExchange: The Changing Landscape for Patent Remedies, 9 Minn. J.L. Sci. & Tech. 543, 572 (2008) (“For the first time, courts are not granting permanent injunctions to many successful patent plaintiffs.”); Robin M. Davis, Failed Attempts to Dwarf the Patent Trolls: Permanent Injunctions in Patent Infringement Cases Under the Proposed Patent Reform Act of 2005 and eBay v. MercExchange, 17 Cornell J.L. & Pub. Pol’y 431, 444 (2008) (“However, the first few district courts deciding patent cases following that decision granted injunctions to patent owners in the majority of cases, at a rate of approximately two-to-one.”).

[40] See generally Epstein & Noroozi, supra note 6, at 1406-08.

[41] Vincenzo Denicolò, Damien Geradin, Anne Layne-Farrar & A. Jorge Padilla, Revisiting Injunctive Relief: Interpreting eBay in High-Tech Industries with Non-Practicing Patent Holders, 4 J. Comp. L. & Econ. 571 (2008).

[42] Id. at 608. See also Vincenzo Denicolò, Do Patents Over-Compensate Innovators?, 22 Econ. Pol’y 681 (2007) (noting that, with respect to patents in general, “a preponderance of what evidence is currently available points against the over-reward hypothesis”).

[43] See, e.g., Mark Schankerman & Suzanne Scotchmer, Damages and Injunctions in Protecting Intellectual Property, 32 RAND J. Econ. 201 (2001).

[44] Epstein & Noroozi, supra note 6, at 1408.

[45] Tsai & Wright, supra note 23, at 163.

^[46] James Ratliff & Daniel L. Rubinfeld, The Use and Threat of Injunctions in the RAND Context, 9 J. Competition L. & Econ. 14 (2013).

^[47] Ratliff & Rubinfeld, supra note 50, at 7 (emphasis added).

[48] Tsai & Wright, supra note 23, at 182.

[49] Call for Comments, supra note 1, Question 9.

[50] See, e.g., Rudi Bekkers, Christian Catalini, Arianna Martinelli, Cesare Righi, and Timothy Simcoe. Disclosure Rules and Declared Essential Patents, 52 Research Policy, 104618, 3 (2023) (“Thus, allowing blanket disclosure can be efficient if the main purpose of a disclosure policy is to reassure prospective implementers that a license will be available. On the other hand, blanket disclosure shifts search costs from the patent holder (who presumably has a comparative advantage at finding its own essential patents) onto other interested parties, such as prospective licensees who wish to evaluate the scope and value of a firm’s dSEPs; other SSO participants seeking to make explicit cost-benefit comparisons of alternative technologies before committing to a standard; and regulators or courts that might use information about relevant dSEPs to determine reasonable royalties.”).

[51] See Jacob & Nikolic, supra note 10.

[52] See European Commission, Impact Assessment Report Accompanying the Document Proposal for a Regulation of the European Parliament and of the Council on Standard Essential Patents and Amending Regulation (EU) 2017/1001, SWD(2023) 124 final (“Impact Assessment”), at 101.

[53] Call for Comments, supra note 1, Question 10.

[54] See, e.g., U.S. Patent Office, Intellectual Property and the U.S. Economy: 2016 Update (2016), available at https://www.uspto.gov/sites/default/files/documents/IPandtheUSEconomySept2016.pdf.

[55] Shayerah Ilias Akhtar, Liana Wong & Ian F. Fergusson, Intellectual Property Rights and International Trade, at 6 (Congressional Research Service, May 12, 2020), available at https://crsreports.congress.gov/product/pdf/RL/RL34292 (“Intellectual property generally is viewed as a long-standing strategic driver of U.S. productivity, economic growth, employment, higher wages, and exports. It also is considered a key source of U.S. comparative advantage, such as in innovation and high-technology products. Nearly every industry depends on it for its businesses. Industries that rely on patent protection include the aerospace, automotive, computer, consumer electronics, pharmaceutical, and semiconductor industries.”).

[56] See, e.g., Martina F. Ferracane & Hosuk Lee-Makiyama, China’s Technology Protectionism and Its Non-negotiable Rationales, ECIPE (Jun. 2017), available at https://ecipe.org/publications/chinas-technology-protectionism. Consider that, even for actual citizens of the People’s Republic of China, individual rights are legally subordinate to “the interests of the state.” Const. of the People’s Rep. of China, Art. 51, available athttp://www.npc.gov.cn/englishnpc/constitution2019/201911/1f65146fb6104dd3a2793875d19b5b29.shtml. One has to imagine that the level of legal protections afforded foreign firms is no better, and surely must be subordinate to the objectives of China’s industrial policy, including the goal of leapfrogging the United States in IP production. See, e.g., Karen M. Sutter, “Made in China 2025” Industrial Policies: Issues for Congress (Congressional Research Service, Aug. 11, 2020), available at https://sgp.fas.org/crs/row/IF10964.pdf.

[57] See China Is Becoming More Assertive in International Legal Disputes, The Economist (Sep. 18, 2021), https://www.economist.com/china/2021/09/18/china-is-becoming-more-assertive-in-international-legal-disputes (“In the past year Chinese courts have issued sweeping orders on behalf of Chinese smartphone-makers that seek to prevent lawsuits against them in other countries over the use of foreign companies’ intellectual property… so that they (rather than foreign courts) can decide how much Chinese firms should pay in royalties to the holders of patents that their products use.”).

[58] See Matthew Laight, Shifting landscape in SEP FRAND litigation – 2021 will see hard fought disputes in China and India, digital business (Dec. 9, 2020), https://digitalbusiness.law/2020/12/shifting-landscape-in-sep-frand-litigation-2021-will-see-hard-fought-disputes-in-china-and-india.

[59] See RPX Corporation, China: Chinese Courts Can Set Global SEP License Terms, Rules Supreme People’s Court, Mondaq (Oct. 21, 2021), https://www.mondaq.com/china/patent/1120114/chinese-courts-can-set-global-sep-license-terms-rules-supreme-people39s-court.

[60] Id.

[61] See Rush Doshi, Emily De La Bruye?re, Nathan Picarsic, & John Ferguson, China as a “Cyber Great Power”: Beijing’s Two Voices in Telecommunications, Brookings Institute Foreign Policy Paper (Apr. 2021) at 16, available at https://www.brookings.edu/wp-content/uploads/2021/04/FP_20210405_china_cyber_power.pdf. (“In March 2018, Beijing launched the China Standards 2035 project, led by the Chinese Academy of Engineering. After a two-year research phase, that project evolved into the National Standardization Development Strategy Research in January 2020. The ‘Main Points of Standardization Work in 2020’ issued by China’s National Standardization Committee in March 2020 outlined intentions to ‘strengthen the interaction between the standardization strategy and major national strategies.’”).

[62] Quoted in id.

[63] Id. “The phrase ‘cyber great power’ is a key concept guiding Chinese strategy in telecommunications as well as IT more broadly. It appears in the title of almost every major speech by President Xi Jinping on China’s telecommunications and network strategy aimed at a domestic audience since 2014. But the phrase is rarely found in messaging aimed at external foreign audiences, appearing only once in six years of remarks by Foreign Ministry spokespersons. This suggests that Beijing intentionally dilutes discussions of its ambitions in order not to alarm foreign audiences.” Id. at 3 (emphasis added).

[64] See Danny Russel & Blake Berger, Is China Stacking the Technology Deck by Setting International Standards?, The Diplomat (Dec. 2, 2021), https://thediplomat.com/2021/12/is-china-stacking-the-technology-deck-by-setting-international-standards.

[65] Emily de la Bruyere, China’s Quest to Shape the World Through Standards Setting, Hinrich Foundation (Jul. 2021), at 11 (emphasis added), available at https://www.hinrichfoundation.com/research/article/tech/china-quest-to-shape-the-world-through-standards-setting.

[66] Although China is currently under-represented in most SDOs, that is already rapidly changing. See Justus Baron & Olia Kanevskaia, Global Competition for Leadership Positions in Standards Development Organizations, Working Paper (Mar. 31, 2021), available at https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3818143. As Baron and Kanevskaia note, “[t]he surge in the number of leadership positions held by Huawei… [have] raised concerns that… Huawei [may] gain an undue competitive advantage over Western commercial and strategic interests.” Id. at 2.

[67] Updegrove, supra note 19.

[68] See id. at 30-36 (surveying the economic benefits from standardization). See also Soon-Yong Choi & Andrew B. Whinston, Benefits and Requirements for Interoperability in the Electronic Marketplace, 2 Tech. in Soc’y 33, 33 (2000) (“Economic benefits of interoperability result in lowered production or transaction costs typically utilizing standardized parts or automated processes. In the networked economy, the need for interoperability extends into an entire commercial processes, market organizations and products.”).

[69] Anna Gross, Madhumita Murgia & Yuan Yang, Chinese Tech Groups Shaping UN Facial Recognition Standards, Financial Times (Dec. 1, 2019), https://www.ft.com/content/c3555a3c-0d3e-11ea-b2d6-9bf4d1957a67 (“‘The drive to shape international standards… reflects longstanding concerns that Chinese representatives were not at the table to help set the rules of the game for the global Internet,’ the authors of the New America report wrote. ‘The Chinese government wants to make sure that this does not happen in other ICT spheres, now that China has become a technology power with a sizeable market and leading technology companies, including in AI.’”).

[70] Russel & Berger, supra note 67.

[71] Request for Comments, supra note 1,  Question 11.

[72] Igor Nikolic, Licensing Negotiations Groups for SEPs: Collusive Technology Buyers Arrangements? Their Pitfalls and Reasonable Alternatives, Les Nouvelles 350 (2021).

[73] Hector Axel Contreras & Julia Brito, Patent Pools: A Practical Perspective – Part II, Les Nouvelles 39 (2022).

[74] Unwired Planet v Huawei [2017] EWHC 711 (Pat).

[75] TCL v Ericsson, Case No. 8:14-cv-003410JVS-DFM (C.D. Cal. 2018); TCL v Ericsson, 943 F.3d 1360 (Fed. Cir. 2019)

[76] Interdigital v Lenovo [2023] EWHC 539 (Pat) 733.

Here in New Jersey, where I live, the day before Halloween is commonly celebrated as “Mischief Night,” an evening of adolescent revelry and light vandalism that typically includes hurling copious quantities of eggs and toilet paper.

It is perhaps fitting, therefore, that President Joe Biden chose Oct. 30 to sign a sweeping executive order (EO) that could itself do quite a bit of mischief. And befitting the Halloween season, in proposing this broad oversight regime, the administration appears to be positively spooked by the development of artificial intelligence (AI).

The order, of course, embodies the emerging and now pervasive sense among policymakers that they should “do something” about AI; the EO goes so far as to declare that the administration feels “compelled” to act on AI. It largely directs various agencies to each determine how they should be involved in regulating AI, but some provisions go further than that. In particular, directives that set new reporting requirements—while ostensibly intended to forward the reasonable goal of transparency—could end up doing more harm than good.

Read the full piece here.

Introduction

We thank you for the opportunity to comment on this important notice of inquiry (NOI)[1] on artificial intelligence (AI) and copyright. We appreciate the U.S. Copyright Office undertaking a comprehensive review of the policy and copyright-law issues raised by recent advances in generative AI systems. This NOI covers key areas that require attention, from legal questions regarding infringement and fair use, to questions about how policy choices could shape opportunities for creators and AI producers to engage in licensing.

At this early date, AI systems have already generated some incredible visual art and impressive written texts, as well as a good deal of controversy. Some artists have banded together as part of an anti-AI campaign;^[2] lawsuits have been filed;^[3] and policy experts have attempted to think through the various legal questions raised by these machine-learning systems.

The debates over the role of AI in creative industries have particular salience for intellectual-property rights. Copyright is notoriously difficult to protect online, and the emergence of AI may exacerbate that difficulty. AI systems also potentially pose an additional wrinkle: it is at least arguable that the outputs they produce can themselves be considered unique creations. There are, of course, other open questions whose answers are relevant here, not the least being whether it is fair to assert that only a human can be “creative” (at least, so far).[4]

But leaving these questions aside, we can say that at least some AI systems produce unique outputs and are not merely routinely duplicating other pieces of work in a digital equivalent of collage. That is, at some level, the machines are engaged in a rudimentary sort of “learning” about how humans arrange creative inputs when generating images, music, or written works. The machines appear to be able to reconstruct this process and produce new sets of words, sounds, or lines and colors that conform to the patterns found in human art, in at least a simulacrum of “creativity.”

But that conclusion isn’t the end of the story. Even if some of these AI outputs are unique and noninfringing, the way that AI systems learn—by ingesting massive quantities of existing creative work—raises a number of thorny copyright-law issues. Indeed, some argue that these systems inherently infringe copyright during the learning phase and that, as discussed below, such processes may not survive a “fair use” analysis.

But nor is that assertion the end of the analysis. Rather, it raises the question of whether applying existing doctrine in this novel technological context yields the best results for society. Moreover, it heightens the need for a comprehensive analytical framework to help parse these questions.

A.            The Law & Economics of Copyright and AI

Nearly all would agree that it is crucial that law and public policy strike the appropriate balance between protecting creators’ existing rights and enabling society to enjoy the potentially significant benefits that could arise from the development of AI systems. Indeed, the subject is often cast as a dramatic conflict between creative professionals struggling to make ends meet and innovative firms working to provide cutting-edge AI technology. For the moment, however, it is likely more important to determine the right questions to ask and the proper analytical framework to employ than it is to identify any precise balancing point.

What is important to remember is that copyright policy is foremost economic in nature and “can be explained as a means for promoting efficient allocation of resources.”[5] That is to say, the reason that property rights in creative expression exist is to guarantee the continued production of such works.[6] The fundamental tradeoff in copyright policy is between the costs of limiting access to creative works, and the value obtained by encouraging production of such works.[7] The same applies in the context of AI: identifying the key tradeoffs and weighing the costs and benefits of restricting access to protected works by the producers (and users) of AI systems.[8]

This entails examining the costs and benefits of relatively stronger or weaker forms copyright protection in terms of their effects on both incentives and access, and as they relate to both copyright holders and AI-system developers. It also requires considering where the transaction costs should be allocated for negotiating access to both copyright and, as discussed infra,[9] the use of name/image/likeness, as well as how those allocations are likely to shape outcomes.

At root, these questions center on how to think about the property rights that limit access to protected works and, possibly even more importantly, how to assign new property rights governing the ability to control the use of a name/image/likeness. As we know from the work of the late Nobel laureate Ronald Coase, the actual demarcation of rights affects parties’ abilities to negotiate superior solutions.^[10] The development of nuisance law provides a good example of the problem at hand. When a legal regime provides either strict liability or no-liability rules around pollution, parties have little incentive to minimize harmful conduct:

The factory that has the absolute right to pollute will, if transaction costs are prohibitive, have no incentives to stop (or reduce) pollution even if the cost of stopping would be much less than the cost of pollution to the homeowners. Conversely, homeowners who have an absolute right to be free from pollution will, if transaction costs are prohibitive, have no incentive to take steps of their own to reduce the effects of pollution even if the cost to them of doing so (perhaps by moving away) is less than the cost to the factory of not polluting or of polluting less.^[11]

As Coase observed, this class of problem is best regarded as reciprocal in nature, and the allocation of rights matters in obtaining an efficient outcome. This is necessarily so because, when fully considered, B’s ability to restrain A from the pollution-generating activity can itself be conceived of as another kind of harm that B can impose on A. Therefore, the problem requires a balancing of the relative harms generated by both A and B in exercising conflicting claims in a particular context.

When thinking about how to minimize harms—whether from pollution or other activity that generates social costs (which is to say, nearly every activity)—the aim is to decide whether “the gain from preventing the harm is greater than the loss which would be suffered elsewhere as a result of stopping the action which produces the harm.”^[12] Theoretically, in a world without transaction costs, even assignments of no-liability or strict-liability rules could be bargained around. But we do not live in such a world.^[13] Thus, “[i]n a world in which there are costs of rearranging the rights established by the legal system [common law and statutory assignments of liability] are, in effect, making a decision on the economic problem and determining how resources are to be employed.”^[14]

While pollution rules, unlicensed uses of intellectual property, and a host of other activities subject to legal sanction are not typically framed as resource-allocation decisions, it is undeniable that they do have this character. This is true even where legislation attempts to correct deficiencies in the system. We experience a form of blindness when we focus on correcting what may be rightly perceived as problems in a liability regime. Such analysis tends to concentrate attention on particular deficiencies of the system and to nourish the belief that any measure that removes the deficiency is necessarily desirable. It diverts attention from other changes inevitably associated with the corrective measure—changes that may well produce more harm than the original deficiency.^[15]

All of this is to say that one solution to the costs generated by the need for AI systems to process a massive corpus of expensive, copyright-protected material is neither to undermine property rights, nor to make AI impossible, but to think about how new property rights could make the system work. It may be that some entirely different form or allocation of property right would facilitate bargaining between rightsholders and AI creators, optimizing resource allocation in a way the existing doctrinal regime may not be able to.

A number of other questions flow from this insight into the allocative nature of copyright. How would the incentives for human creators change under different copyright rules for AI systems, or in the face of additional rights? And how would access to copyrighted works for AI training change with different rules, and what effects would that access have on AI innovation?

Above all, our goal today should be to properly frame the AI and copyright debate by identifying tradeoffs, quantifying effects (where possible), and asking what rules best serve the overall objectives of the copyright system and the social goal of encouraging AI innovation. The best chance of striking the right balance will come from a rigorous framing of the questions and from the use of economic analysis to try to answer them.

B.            Copyright Law and AI: Moving Forward

As the Copyright Office undertakes this inquiry, it is important to recognize that, regardless of how the immediate legal questions around AI and copyright are resolved, the growing capabilities and adoption of generative AI systems will likely necessitate some changes in the long term.

The complex questions surrounding the intersection of AI and copyright law admit reasonable arguments on both sides. But AI is here to stay, regardless, and if copyright law is applied in an unduly restrictive manner that substantially hinders socially beneficial AI innovation, it could provoke a broader public-policy backlash that does more to harm copyright’s ability to protect creative works than it does to stanch AI’s ability to undermine it. Copyright law risks being perceived as an obstruction to technological progress if it is used preemptively to kill AI in the cradle. Such an outcome could galvanize calls for recalibrating copyright’s scope and protections in the name of the public interest.

This illustrates the precarious balancing act that copyright law faces in the wake of rapidly evolving technologies like AI. Aggressive copyright restrictions that curtail AI development could instigate a public-policy counter-reaction before Congress and the courts that ultimately undermines copyright’s objectives. The judicious course is to adapt copyright law cautiously to enable AI’s responsible evolution, while resolutely preserving the incentives for human creativity.

In the remainder of this analysis, we offer our perspective on the likely outcomes of the AI-copyright issues raised in this NOI, given the current state of the law. These assessments reflect our perspective formed through the rigorous application of established copyright principles and precedent to the novel technological context of generative AI systems. Reasonable arguments rooted in existing doctrine could be made to support different conclusions. We submit these comments not as definitive predictions or normative preferences, but rather as informed appraisals of how courts may analyze AI under present copyright law, absent legislative intervention.

We appreciate the Copyright Office starting this process to modernize copyright law for the AI age. This inquiry is an important first step, but openness to further evolution will be key to promoting progress in both AI and the arts. We believe an open, evidence-based discussion of these issues will lead to balanced solutions that uphold copyright’s constitutionally mandated purpose, while allowing responsible AI innovation for the public benefit.

II.            The Training of AI Systems and the Applicability of Fair Use

In the NOI, the Copyright Offices asks: “[u]nder what circumstances would the unauthorized use of copyrighted works to train AI models constitute fair use?”^[16]

To answer this question, it would be useful to first briefly walk through a high-level example of how AI systems work, in order to address the most relevant points of contact between AI systems and copyright law.

A.            A Brief Technical Description of AI Training

AI-generated content is not a single “thing,” but a collection of differing processes, each with different implications for the law. For the purposes of this discussion, we will discuss image generation using “generated adversarial networks” (GANs) and diffusion models. Although different systems and different types of content generation will vary, the basic concepts discussed below are nonetheless useful at a general level.^[17]

A GAN is a type of machine-learning model that consists of two parts: a generator and a discriminator.^[18] The generator is trained to create new images that look like they come from a particular dataset, while the discriminator is trained to distinguish the generated images from real images in its original dataset.^[19] The two parts are trained together in an adversarial manner, with the generator trying to produce images that can fool the discriminator and the discriminator trying to correctly identify the generated images.^[20]

A diffusion model, by contrast, analyzes the distribution of information in an image, as noise is progressively added to it.^[21] This kind of algorithm analyzes characteristics of sample images, like the distribution of colors or lines, in order to understand what counts as an accurate representation of a subject (i.e., what makes a picture of a cat look like a cat, and not like a dog).^[22]

For example, in the generation phase, diffusion-based systems start with randomly generated noise, and work backward in “denoising” steps to essentially “see” shapes:

The sampled noise is predicted so that if we subtract it from the image, we get an image that’s closer to the images the model was trained on (not the exact images themselves, but the distribution – the world of pixel arrangements where the sky is usually blue and above the ground, people have two eyes, cats look a certain way – pointy ears and clearly unimpressed).^[23]

While it is possible that some implementations might be designed in a way that saves copies of the training images,^[24] for at least some systems, once the network is trained using these techniques, it will not need to rely on saved copies of input work in order to produce outputs. The models that are produced during training are, in essence, instructions to a different piece of software about how to start with a prompt from a user, a palette of pure noise, and progressively “discover” signal in that image until some new image emerges.

B.            Fair Use

The creator of some of the most popular AI tools, OpenAI, is not shy about their use of protected works in the training phase of the algorithms. In comments to the U.S. Patent and Trademark Office (PTO), OpenAI noted that:

Modern AI systems require large amounts of data. For certain tasks, that data is derived from existing publicly accessible “corpora”… of data that include copyrighted works. By analyzing large corpora (which necessarily involves first making copies of the data to be analyzed), AI systems can learn patterns inherent in human-generated data and then use those patterns to synthesize similar data which yield increasingly compelling novel media in modalities as diverse as text, image, and audio. (emphasis added).^[25]

Thus, at the training stage, the most popular forms of AI systems require making copies of existing works. And where that material is either not in the public domain or is not licensed, an infringement can occur. Thus, the copy must not be infringing (say, because it is transient), or some affirmative defense is needed to excuse the infringement. Toward this end, OpenAI believes that this use should qualify as fair use,^[26] as do most or all the other major producers of generative AI systems.^[27]

But as OpenAI has framed the fair-use analysis, it is not clear that these uses should qualify. There are two major questions in this respect: will the data used to train these systems count as “copies” under the Copyright Act, and, if so, is the use of these “copies” sufficiently “transformative” to qualify for the fair-use defense?

1.              Are AI systems being trained with ‘copies’ of protected works?

Section 106 of the Copyright Act grants the owner of a copyright the exclusive right “to reproduce… copyrighted work in copies” and to authorize others to do so.^[28] If an AI system makes a copy of a file to a computer during training, this would likely constitute a prima facie violation of the copyright owner’s exclusive right of reproduction under Section 106. This is fairly straightforward.

But what if the “copy” is “transient” and/or only partial pieces of content are used in the training? For example, what if a training program merely streamed small bits of a protected work into temporary memory as part of its training, and retained no permanent copy?

As the Copyright Office has previously observed, even temporary reproductions of a work in a computer’s memory can constitute “copies” under the Copyright Act.^[29] Critically, this includes even temporary reproductions made as part of a packet-switching network transmission, where a particular file is broken into individual packets, because the packets can be reassembled into substantial portions or even entire works.^[30] On the topic of network-based transmission, the Copyright Office further observed that:

Digital networks permit a single disk copy of a work to meet the demands of many users by creating multiple RAM copies. These copies need exist only long enough to be perceived (e.g., displayed on the screen or played through speakers), reproduced or otherwise communicated (e.g., to a computer’s processing unit) in order for their economic value to be realized. If the network is sufficiently reliable, users have no need to retain copies of the material. Commercial exploitation in a network environment can be said to be based on selling a right to perceive temporary reproductions of works.^[31]

This is a critical insight that translates well to the context of AI training. The “transience” of the copy matters with respect to the receiver’s ability to perceive the work in a way that yields commercial value. Under this reasoning, the relevant locus of analysis is on the AI system’s ability to “perceive” a work for the purposes of being trained to “understand” the work. In this sense, you could theoretically find the existence of even more temporary copies than that necessary for human perception to implicate the reproduction right.

Even where courts have been skeptical of extending the definition of “copy” to “fleeting” copies in computer memory, this underlying logic is revealed. In Cartoon Network LP, LLLP v. CSC Holdings, Inc., 536 F.3d 121 (2008), the 2nd U.S. Circuit Court of Appeals had to determine whether buffered media sent to a DVR device was too “transient” to count as a “copy”:

No bit of data remains in any buffer for more than a fleeting 1.2 seconds. And unlike the data in cases like MAI Systems, which remained embodied in the computer’s RAM memory until the user turned the computer off, each bit of data here is rapidly and automatically overwritten as soon as it is processed. While our inquiry is necessarily fact-specific, and other factors not present here may alter the duration analysis significantly, these facts strongly suggest that the works in this case are embodied in the buffer for only a “transitory” period, thus failing the duration requirement.^[32]

In Cartoon Network, the court acknowledged both that the duration analysis was fact-bound, and also that the “fleeting” nature of the reproduction was important. “Fleeting” is a relative term, based on the receiver’s capacities. A ball flying through the air may look “fleeting” to a human observer, but may appear to go much more cognizable to a creature with faster reaction time, such as a house fly. So, too, with copies of a work in a computer’s memory and the ability to “perceive” what is fixed in a buffer: what may be much too quick for a human to perceive may very well be within an AI system’s perceptual capabilities.

Therefore, however the training copies are held, there is a strong possibility that a court will find them to be “copies” for the purposes of the reproduction right—even with respect to partial copies that exist for very small amounts of time.

2.              The purpose and character of using protected works to train AI systems

Fair use provides for an affirmative defense against infringement when the use is, among other things, “for purposes such as criticism, comment, news reporting, teaching…, scholarship, or research.”^[33] When deciding whether a fair-use defense is applicable, a court must balance a number of factors:

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.^[34]

The fair-use defense that AI creators have advanced is rooted in the first factor: the nature and character of the use. Although a full analysis of all the factors is ultimately necessary, analysis of the first factor is sufficiently complicated to warrant full attention here. In particular, the complex issue at hand is whether uses of protected works to train AI systems are sufficiently “transformative” or not.^[35]

Whether the use of a copyrighted work to train an AI is “transformative” is certainly a novel question, but it is one that will likely be answered in light of an observation the U.S. Supreme Court made in Campbell v. Acuff Rose Music:

[W]hen a commercial use amounts to mere duplication of the entirety of an original, it clearly “supersede[s] the objects,”… of the original and serves as a market replacement for it, making it likely that cognizable market harm to the original will occur… But when, on the contrary, the second use is transformative, market substitution is at least less certain, and market harm may not be so readily inferred.^[36]

Moreover, “[t]he word ‘transformative’ cannot be taken too literally as a sufficient key to understanding the elements of fair use. It is rather a suggestive symbol for a complex thought, and does not mean that any and all changes made to an author’s original text will necessarily support a finding of fair use.”[37] A key question, then, is whether training AI systems on copyrighted works amounts to a mere “duplication of the entirety of an original” or is sufficiently “transformative” to support a fair-use defense. As noted above, OpenAI believes that its use is transformative. According to its comments:

Training of AI systems is clearly highly transformative. Works in training corpora were meant primarily for human consumption for their standalone entertainment value. The “object of the original creation,” in other words, is direct human consumption of the author’s ?expression.? Intermediate copying of works in training AI systems is, by contrast, “non-expressive” the copying helps computer programs learn the patterns inherent in human-generated media. The aim of this process—creation of a useful generative AI system—is quite different than the original object of human consumption. The output is different too: nobody looking to read a specific webpage contained in the corpus used to train an AI system can do so by studying the AI system or its outputs. The new purpose and expression are thus both highly transformative.^[38]

This framing, however, works against OpenAI’s interests. As noted above, and reinforced in the immediately preceding quote, generative AI systems are made of at least two distinct pieces. The first is a piece of software that ingests existing works and creates a file that can serve as instructions to the second piece of software. The second piece of software takes the output of the first and can produce independent results. Thus, there is a clear discontinuity in the process whereby the ultimate work created by the system is disconnected from the creative inputs used to train the software.

Therefore, the protected works are arguably ingested into the first part of the system “for their standalone entertainment value.” That is to say, the goal of copying and showing a protected work to an AI system is for the analog of “direct human consumption of the author’s expression” in order for the system to learn about that expression.

The software is learning what counts as “standalone entertainment value” and therefore the works must be used in those terms. Surely, a computer is not sitting on a couch and surfing for its pleasure. But it is solely for the very “standalone entertainment value” that the first piece of software is being shown copyrighted material. By contrast, parody or “remixing” uses incorporate a work into some secondary expression that directly transforms the input. The way these systems work is to learn what makes a piece entertaining and then to discard that piece altogether. Moreover, this use for the art qua art most certainly interferes with the existing market, insofar as this use is in lieu of reaching a licensing agreement with rightsholders.

A good analogy is art students and art textbooks. Art students view protected works in an art textbook in order to learn how to reproduce the styles contained therein. The students would not be forgiven for pirating the textbooks merely because they intend to go on to make new paintings. They would still be liable for copyright infringement if they used unlicensed protected works as part of their education.

The 2nd U.S. Circuit Court of Appeals dealt with a case that operates similarly to this dynamic. In American Geophysical Union v. Texaco, 60 F.3d 913 (2d Cir. 1994), the 2nd Circuit considered whether Texaco’s photocopying of scientific articles produced by the plaintiffs qualified for a fair-use defense. Texaco employed between 400 and 500 research scientists and, as part of supporting their work, maintained subscriptions to a number of scientific journals.^[39]

It was common practice for Texaco’s scientists to photocopy entire articles and save them in a file.^[40] The plaintiffs sued for copyright infringement.^[41] Texaco asserted that photocopying by its scientists for the purposes of furthering scientific research—that is to train the scientists on the content of the journal articles—should count as a fair use. The argument was, at least in part, that this was sufficiently “transformative,” because the scientists were using that knowledge to invent new products.^[42] The 2nd Circuit disagreed:

The “transformative use” concept is pertinent to a court’s investigation under the first factor because it assesses the value generated by the secondary use and the means by which such value is generated. To the extent that the secondary use involves merely an untransformed duplication, the value generated by the secondary use is little or nothing more than the value that inheres in the original. Rather than making some contribution of new intellectual value and thereby fostering the advancement of the arts and sciences, an untransformed copy is likely to be used simply for the same intrinsic purpose as the original, thereby providing limited justification for a finding of fair use….^[43]

The 2nd Circuit thus observed that copies of the scientific articles were made solely to consume the material itself. AI developers often make an argument analogous to that made by Texaco: that training AI systems surely advances scientific research, and therefore fosters the “advancement of the arts and sciences.” But in American Geophysical Union, the initial copying of copyrighted content, even where it was ultimately used for the “advancement of the arts and sciences,” was not held to be sufficiently “transformative.”^[44] The case thus stands for the proposition that one cannot merely identify a social goal down that would be advanced at some future date in order to permit an exception to copyright protection. As the court put it:

[T]he dominant purpose of the use is a systematic institutional policy of multiplying the available number of copies of pertinent copyrighted articles by circulating the journals among employed scientists for them to make copies, thereby serving the same purpose for which additional subscriptions are normally sold, or… for which photocopying licenses may be obtained.[45]

The use itself must be transformative and different, and copying is not transformative merely because it may be used as an input into a later transformative use. By the same token, therefore, it seems likely that where an AI system ingests (copies) copyrighted works, that use is similarly not transformative, despite its ultimate use as an input in the creation of other original works.

Comparing the American Geophysical Union analysis with the search-engine “snippets” and “thumbnails” cases provides a useful comparison relevant to the AI analysis. In Kelly v. Arriba Soft Corp., 336 F.3d 811 (9^th Cir. 2002), the 9^th U.S. Circuit Court of Appeals ruled that a search engine’s creation of thumbnail images from original copies was a transformative fair use.[46] Arriba’s search-engine crawler made full-sized copies of Kelly’s images and stored them temporarily on Arriba’s server to generate thumbnail versions. After the thumbnails were created, the full-sized originals were deleted. The thumbnails were used to facilitate Arriba’s image-based search engine. In reaching its fair-use conclusion, the 9^th Circuit opined that:

Arriba’s use of Kelly’s images promotes the goals of the Copyright Act and the fair use exception. The thumbnails do not stifle artistic creativity because they are not used for illustrative or artistic purposes and therefore do not supplant the need for the originals.[47]

Further, although “Arriba made exact replications of Kelly’s images, the thumbnails were much smaller, lower-resolution images that served an entirely different function than Kelly’s original images.”[48]

The court found it important that the search engine did not use the protected works for their intended “aesthetic experience,” but rather for the purpose of constructing a search index.[49] Indeed, the entire point of a search engine is not to “supersede” the original, but in many or most cases to provider users an efficient means to find that original online.[50]

The court discussed, but only briefly, the benefit to the public of Arriba’s transformative use,[51] noting that “[Arriba’s thumbnails] benefit the public by enhancing information-gathering techniques on the internet.”[52] Five years later, in Perfect 10 Inc. v. Amazon.com Inc., 487 F.3d 701 (2007), the 9^th Circuit expanded on this question somewhat.[53] There, in holding that the novelty of the use was of crucial importance to the analysis,[54] the court also stressed that the value of that use was a function of its newness:

[A] search engine provides social benefit by incorporating an original work into a new work, namely, an electronic reference tool. Indeed, a search engine may be more transformative than a parody [the use at issue in Campbell] because a search engine provides an entirely new use for the original work, while a parody typically has the same entertainment purpose as the original work.[55]

Indeed, even in light of the commercial nature of Google’s use of copyrighted content in its search engine, its significant public benefit carried the day: “We conclude that the significantly transformative nature of Google’s search engine, particularly in light of its public benefit, outweighs Google’s superseding and commercial uses of the thumbnails in this case.”[56] And, of particular relevance to these questions in the context of AI, the court in Perfect 10 went on to “note the importance of analyzing fair use flexibly in light of new circumstances.”[57]

Ultimately, the Perfect 10 decision tracked Kelly fairly closely on the rest of the “transformativeness” analysis in finding fair use, because “[a]lthough an image may have been created originally to serve an entertainment, aesthetic, or informative function, a search engine transforms the image into a pointer directing a user to a source of information.”[58]

The core throughline in this line of cases is the question of whether a piece of content is being used for its expressive content, weighed against the backdrop of whether the use is for some new (and, thus, presumptively valuable) purpose. In Perfect 10 and Kelly, the transformative use was the creation of a search index.

“Snippets” fair-use cases track a similar line of reasoning. For example, in Authors Guild v. Google Inc., 804 F.3d 202 (2d Cir. 2015), the 2nd Circuit ruled that Google’s use of “snippets” of copyrighted books in its Library Project and Google Books website was a “transformative” fair use.[59] Holding that the “snippet view” of books digitized as part of the Google Books project did not constitute an effectively competing substitute to the original works, the circuit court noted that copying for the purpose of “criticism” or—as in that case—copying for the purpose of “provision of information about” the protected work, “tends most clearly to satisfy Campbell’s notion of the ‘transformative’ purpose.”[60]

Importantly, the court emphasized the importance of the public-benefit aspect of transformative uses: “[T]ransformative uses tend to favor a fair use finding because a transformative use is one that communicates something new and different from the original or expands its utility, thus serving copyright’s overall objective of contributing to public knowledge.”[61]

Underscoring the idea that the “transformativeness” analysis weighs whether a use is merely for expressive content against the novelty/utility of the intended use, the court observed:

Google’s division of the page into tiny snippets is designed to show the searcher just enough context surrounding the searched term to help her evaluate whether the book falls within the scope of her interest (without revealing so much as to threaten the author’s copyright interests). Snippet view thus adds importantly to the highly transformative purpose of identifying books of interest to the searcher.[62]

Thus, the absence of use of the work’s expressive content, coupled with a fairly circumscribed (but highly novel) use was critical to the outcome.

The entwined questions of transformative use and the public benefit it confers are significantly more complicated in the AI context, however. Unlike the incidental copying involved in search-engine indexing or thumbnails, training generative AI systems directly leverages copyrighted works for their expressive value. In the Google Books and Kelly cases, the defendant systems extracted limited portions of works or down-sampled images solely to identify and catalog their location for search purposes. The copies enabled indexing and access, and they expanded public knowledge through a means unrelated to the works’ protected aesthetics.

But in training AI models on copyrighted data, the systems necessarily parse the intrinsic creative expression of those works. The AI engages with the protected aesthetic elements themselves, not just superficial markers (like title, length, location on the internet, etc.), in order to internalize stylistic and compositional principles. This appropriates the heart of the works’ copyright protection for expressive ends, unlike the more tenuous connections in search systems.

The AI is thus “learning” directly from the protected expression in a manner akin to a human student studying an art textbook, or like the scientists learning from the journals in American Geophysical Union. The subsequent AI generations are built from mastery of the copyrighted training materials’ creative expression. Thus, while search-engine copies only incidentally interact with protected expression to enable unrelated innovation, AI training is predicated on excavating the protected expression itself to fuel iterative creation. These meaningfully different purposes have significant fair-use implications.

This functional difference is, as noted, central to the analysis of a use’s “purpose and character.” Indeed, “even making an exact copy of a work may be transformative so long as the copy serves a different function than the original work.”[63] But the benefit to the public from the new use is important, as well, particularly with respect to the possible legislative response that a restrictive interpretation of existing doctrine may engender.

If existing fair-use principles prohibit the copying required for AI, absent costly item-by-item negotiation and licensing, the transaction costs could become prohibitive, thwarting the development of technologies that promise great public value.[64] Copyright law has faced similar dilemmas before, where the transaction costs of obtaining permission for socially beneficial uses could frustrate those uses entirely.[65] In such cases, we have developed mechanisms like compulsory licensing to facilitate the necessary copying, while still attempting to compensate rightsholders. An unduly narrow fair-use finding for AI training could spur calls for similar interventions in service of enabling AI progress.

In other words, regardless of the veracity of the above conclusion that AI’s use of copyrighted works may not, in fact, serve a different function than the original, courts and legislators may be reluctant to allow copyright doctrine to serve as an absolute bar against self-evidently valuable activity like AI development. Our aim should be to interpret or recalibrate copyright law to permit such progress while upholding critical incentives for creators.

C.            Opt-In vs. Opt-Out Use of Protected Works

The question at the heart of the prior discussion—and, indeed, at the heart of the economic analysis of copyright—is whether the transaction costs that accompany requiring express ex ante permission for the use of protected works are so high that they impedes socially beneficial conduct whose value would outweigh the social cost of allowing permissionless and/or uncompensated use.[66] The NOI alludes to this question when it asks: “Should copyright owners have to affirmatively consent (opt in) to the use of their works for training materials, or should they be provided with the means to object (opt out)?”^[67]

This is a complex problem. Given the foregoing thoughts on fair use, it seems quite possible that, at present, the law requires creators of AI systems to seek licenses for protected content, or else must resort to public-domain works for training. Given the volume of copyrighted works that AI developers currently use to train these systems, such requirements may be broadly infeasible.

On one hand, requiring affirmative opt-in consent from copyright holders imposes significant transaction costs on AI-system developers to identify and negotiate licenses for the vast amounts of training data required. This could hamper innovation in socially beneficial AI systems. On the other hand, an opt-out approach shifts more of the transaction-cost burden to copyright holders, who must monitor and object to unwanted uses of their works. This raises concerns about uncompensated use.

Ultimately, the question is where the burden should lie: with AI-system developers to obtain express consent, or with copyright holders to monitor and object to uses? Requiring some form of consent may be necessary to respect copyright interests. Yet an opt-out approach may strike the right balance, by shifting some of the burden back to AI developers while avoiding the infeasibly high transaction costs of mandatory opt-in consent. The optimal approach likely involves nuanced policymaking to balance these competing considerations. Moreover, as we discuss infra, the realistic outcome is most likely going to require rethinking the allocation of property rights in ways that provide for large-scale licensing. Ideally, this could be done through collective negotiation, but perhaps at a de minimis rate, while allowing creators to bargain for remuneration on the basis of other rights, like a right of publicity or other rights attached to the output of AI systems, rather than the inputs.[68]

1.              Creator consent

Relatedly, the Copyright Office asks: “If copyright owners’ consent is required to train generative AI models, how can or should licenses be obtained?”^[69]

Licensing markets exist, and it is entirely possible that major AI developers and large groups of rightsholders can come to mutually beneficial terms that permit a sufficiently large body of protected works to be made available as training data. Something like a licensing agency for creators who choose to make their works available could arise, similar to the services that exist to provide licensed music and footage for video creators.^[70] It is also possible for some to form collective-licensing organizations to negotiate blanket permissions covering many works.

It’s important to remember that our current thinking is constrained by our past experience. All we know today are AI models trained on vast amounts of unlicensed works. It is entirely possible that, if firms were required to seek licenses, unexpected business models would emerge to satisfy both sides of the equation.

For example, an AI firm could develop its own version of YouTube’s ContentID, which would allow creators to control when their work is used in AI training. For some well-known artists, this could be negotiated with an upfront licensing fee. On the user side, any artist who has opted in could then be selected as a “style” for the AI to emulate—triggering a royalty payment to the artist when a user generates an image or song in that style. Creators could also have the option of removing their influence from the system if they so desire.

Undoubtedly, there are other ways to structure the relationship between creators and AI systems  that would facilitate creators’ monetization of the use of their work in AI systems, including legal and commercial structures that create opportunities for both creators and AI firms to succeed.

III.          Generative AI Outputs: Protection of Outputs and Outputs that Infringe

The Copyright Office asks: “Under copyright law, are there circumstances when a human using a generative AI system should be considered the ‘author’ of material produced by the system?”^[71]

Generally speaking, we see no reason why copyright law should be altered to afford protection to purely automatic creations generated by AI systems. That said, when a human makes a nontrivial contribution to generative AI output—such as editing, reframing, or embedding the AI-generated component within a larger work—the resulting work should qualify for copyright protection.

Copyright law centers on the concept of original human authorship.^[72] The U.S. Constitution expressly limits copyright to “authors.”^[73] As of this writing, however, generative AI’s capacities do not rise to the level of true independent authorship. AI systems remain tools that require human direction and judgment.[74] As such, when a person provides the initial prompt or framing, makes choices regarding the iterative development of the AI output, and decides that the result is satisfactory for inclusion in a final work, they are fundamentally engaging in creative decision making that constitutes authorship under copyright law.

As Joshua Gans has observed of recent Copyright Review Board decisions:

Trying to draw some line between AI and humans with the current technology opens up a massive can of worms. There is literally no piece of digital work these days that does not have some AI element to it, and some of these mix and blur the lines in terms of what is creative and what is not. Here are some examples:

A music artist uses AI to denoise a track or to add an instrument or beat to a track or to just get a composition started.

A photographer uses Photoshop or takes pictures with an iPhone that already uses AI to focus the image and to sort a burst of images into one that is appropriate.

A writer uses AI to prompt for some dialogue when stuck at some point or to suggest a frame for writing a story.[75]

Attempting to separate out an “AI portion” from the final work, as the Copyright Review Board proposed, fundamentally misunderstands the integrated nature of the human-AI collaborative process. The AI system cannot function without human input, and its output remains raw material requiring human creativity to incorporate meaningfully into a finished product.

Therefore, when a generative AI system is used as part of a process guided by human creative choices, the final work should be protected by copyright, just as a work created using any other artistic tool or collaborator would be. Attenuating copyrightability due to the use of AI would undermine basic copyright principles and fail to recognize the essentially human nature of the creative process.

A.            AI Outputs and Infringement

The NOI asks: “Is the substantial similarity test adequate to address claims of infringement based on outputs from a generative AI system, or is some other standard appropriate or necessary?” (Question 23)

The outputs of AI systems may or may not violate IP laws, but there is nothing inherent in the processes described above that dictates that they must. As noted, the most common AI systems do not save copies of existing works, but merely “instructions” (more or less) on how to create new work that conforms to patterns found by examining existing work. If we assume that a system isn’t violating copyright at the input stage, it’s entirely possible that it can produce completely new pieces of art that have never before existed and do not violate copyright.

They can, however, be made to violate copyrights. For example, these systems can be instructed to generate art, not just in the style of a particular artist, but art that very closely resembles existing pieces. In this sense, it would be making a copy that theoretically infringes. The fact of an AI’s involvement would not change the analysis: just as with a human-created work, if it is substantially similar to a copyrighted work, it may be found infringing.

There is, however, a common bug in AI systems that leads to outputs that are more likely to violate copyright in this way. Known as “overfitting,” the training leg of these AI systems can be presented with samples that contain too many instances of a particular image.^[76] This leads to a dataset that contains too much information about the specific image, such that—when the AI generates a new image—it is constrained to producing something very close to the original. Similarly, there is evidence that some AI systems are “memorizing” parts of protected books.^[77] This could lead to AI systems repeating copyright-protected written works.

1.              The substantial-similarity test

The substantial-similarity test remains functionally the same when evaluating works generated using AI. To find “substantial similarity,” courts require evidence of copying, as well as an expression that is substantially similar to a protected work.^[78] “It is now an axiom of copyright law that actionable copying can be inferred from the defendant’s access to the copyrighted work and substantial similarity between the copyrighted work and the alleged infringement.”^[79] In many or most cases, it will arguably be the case that AI systems have access to quite a wide array of protected works that are posted online. Thus, there may not be a particularly high hurdle to determine that an AI system actually copied a protected work.

There is, however, one potential problem for the first prong of this analysis. Models produced during a system’s training process do not (usually) contain the original work, but are the “ideas” that the AI systems generated during training. Thus, where the provenance of works contained in a training corpus is difficult to source, it may not be so straightforward to make inferences about whether a model “saw” a particular work. This is because the “ideas” that the AI “learns” from its training corpus are unprotected under U.S. copyright law, as it is permissible to mimic unprotected elements of a copyrighted work (such as ideas).^[80]

Imagine a generative AI system trained on horror fiction. It would be possible for this system to produce a new short story that is similar to one written by Stephen King, but the latent data in the model almost certainly would not violate any copyrights that King holds in his work. The model would contain “ideas” about horror stories, including those learned from an array of authors who were themselves influences on Stephen King, and potentially some of King’s own stories. What the AI system “learns” in this case is the relationship between words and other linguistic particularities that are commonly contained in horror fiction. That is, it has “ideas” about what goes into a horror story, not (theoretically) the text of the horror story itself.

Thus, when demonstrating indirect proof of copying in the case of a Stephen King story, it may pose a difficulty that an AI system has ingested all of H.P. Lovecraft’s work—an author who had a major influence on King. The “ideas” in the model and the output it subsequently produces may, in fact, produce something similar to a Stephen King work, but it may have been constructed largely or entirely on material from Lovecraft and other public-domain horror writers. The problem becomes only more complicated when you realize that this system could also have been trained on public-domain fan fiction written in the style of Stephen King. Thus, for the purposes of the first prong of this analysis, courts may place greater burden on plaintiffs in copyright actions against model producers to demonstrate more than merely that a work was merely available online.

Assuming that plaintiffs are able to satisfy the first prong, once an AI system “expresses” those ideas, that expression could violate copyright law under the second prong of the substantial-similarity test. The second prong inquires whether the final work appropriated the protected original expression.^[81] Any similarities in unprotectable ideas, facts, or common tropes are disregarded.^[82] So, in both traditional and AI contexts, the substantial-similarity test ultimately focuses on the protected components of creative expression, not surface similarity.

The key determination is whether the original work’s protected expression itself has been impermissibly copied, no matter the process that generated the copy. AI is properly viewed as simply another potential tool that could be used in certain acts of copying. It does not require revisiting settled principles of copyright law.

B.            Direct and Secondary Liability

The NOI asks: “If AI-generated material is found to infringe a copyrighted work, who should be directly or secondarily liable—the developer of a generative AI model, the developer of the system incorporating that model, end users of the system, or other parties?”[83]

Applying traditional copyright-infringement frameworks to AI-generated works poses unique challenges in determining direct versus secondary liability. In some cases, the AI system itself may create infringing content without any direct human causation.

1.              Direct liability

If the end user prompts an AI system in a way that intentionally targets copyrighted source material, they may meet the threshold for direct infringement by causing the AI to reproduce protected expression.^[84] Though many AI prompts contain only unprotected ideas, users may sometimes input copyrightable material as the basis for the AI output. For example, a user could upload a copyrighted image and request the AI to make a new drawing based on the sample. In such cases, the user is intentionally targeting copyrighted works and directly “causing” the AI system to reproduce output that is similar. If sufficiently similar, that output could infringe on the protected input. This would be a question of first impression, but it is a plausible reading of available cases.

For example, in CoStar Grp. Inc. v. LoopNet Inc., 373 F.3d 544 (4th Cir. 2004), the 4^th U.S. Circuit Court of Appeals had to consider whether an internet service provider (ISP) could be directly liable when third parties reposted copyrighted material owned by the plaintiff. In determining that merely owning the “machine” through which copies were made or transmitted was not enough to “cause” a direct infringement, the court held that:

[T]o establish direct liability under §§ 501 and 106 of the Act, something more must be shown than mere ownership of a machine used by others to make illegal copies. There must be actual infringing conduct with a nexus sufficiently close and causal to the illegal copying that one could conclude that the machine owner himself trespassed on the exclusive domain of the copyright owner. The Netcom court described this nexus as requiring some aspect of volition or causation… Indeed, counsel for both parties agreed at oral argument that a copy machine owner who makes the machine available to the public to use for copying is not, without more, strictly liable under § 106 for illegal copying by a customer. The ISP in this case is an analogue to the owner of a traditional copying machine whose customers pay a fixed amount per copy and operate the machine themselves to make copies. When a customer duplicates an infringing work, the owner of the copy machine is not considered a direct infringer. Similarly, an ISP who owns an electronic facility that responds automatically to users’ input is not a direct infringer.^[85]

Implied in the 4^th Circuit’s analogy is that, while the owner of a copying machine might not be a direct infringer, a user employing such a machine could be a direct infringer. It’s an imperfect analogy, but a user of an AI system prompting it to create a “substantially similar” reproduction of a protected work could very well be a direct infringer under this framing. Nevertheless, the analogy is inexact, because the user feeds an original into a copying machine in order to make a more-or-less perfect copy of the original, whereas an AI system generates something new but similar. The basic mechanism of using a machine to try to reproduce a protected work, however, remains essentially the same. Whether there is an infringement would be a question of “substantial similarity.”

2.              Secondary liability

As in the case of direct liability, the nature of generative AI makes the secondary-liability determination slightly more complicated, as well. That is, paradoxically, the basis for secondary liability could theoretically arise even where there was no direct infringement.[86]

The first piece of this analysis is relatively easier. If a user is directly liable for infringing a protected work, as noted above, the developer and provider of a generative AI system may face secondary copyright liability. If the AI developer or distributor knows the system can produce infringing outputs, and provides tools or material support that allows users to infringe, it may be liable for contributory infringement.^[87] Critically, merely designing a system that is capable of infringing is not enough to find contributory liability.^[88]

An AI producer or distributor may also have vicarious liability, insofar as it has the right and ability to supervise users’ activity and a direct financial interest in that activity.^[89] AI producers have already demonstrated their ability to control users’ behavior to thwart unwanted uses of the service.^[90] Thus, if there is a direct infringement by a user, a plausible claim for vicarious liability could be made so long as there is sufficient connection between the user’s behavior and the producer’s financial interests.

The question becomes more complicated when a user did not direct the AI system to infringe. When the AI generates infringing content without user direction, it’s not immediately clear who would be liable for the infringement.^[91] Consider the case where, unprompted by either the user or the AI producer, an AI system creates an output that would infringe under the substantial-similarity test. Assuming that the model has not been directed by the producer to “memorize” the works it ingests, the model itself consists of statistical information about the relationship between different kinds of data. The infringer, in a literal sense, is the AI system itself, as it is the creator of the offending output. Technically, this may be a case of vicarious liability, even without an independent human agent causing the direct infringement.

We know that copyright protection can only be granted to humans. As the Copyright Review Board recently found in a case deciding whether AI-generated outputs can be copyrighted:

The Copyright Act protects, and the Office registers, “original works of authorship fixed in any tangible medium of expression.” 17 U.S.C. § 102(a). Courts have interpreted the statutory phrase “works of authorship” to require human creation of the work.^[92]

But can an AI system directly violate copyright? In his Aereo dissent, Justice Clarence Thomas asserted that it was a longstanding feature of copyright law that violation of the performance right required volitional behavior.^[93] But the majority disagreed with him, holding that, by running a fully automated system of antennas intended to allow users to view video at home, the system gave rise to direct copyright liability.^[94] Thus, implied in the majority’s opinion is the idea that direct copyright infringement does not require “volitional” conduct.

It is therefore plausible that a non-sentient, fully automated AI system could infringe copyright, even if, ultimately, there is no way to recover against the nonhuman agent. That does, however, provide an opportunity for claims of vicarious liability against the AI producer or distributor— at least, where the producer has the power to control the AI system’s behavior and that behavior appears to align with the producer’s financial interests.

3.              Protecting the ‘style’ of human creators

The NOI asks: “Are there or should there be protections against an AI system generating outputs that imitate the artistic style of a human creator (such as an AI system producing visual works ‘in the style of’ a specific artist)?”[95]

At the federal level, one candidate for protection against AI imitating some aspects of a creator’s works can currently be found in trademark law. Trademark law, governed by the Lanham Act, protects names, symbols, and other source identifiers that distinguish goods and services in commerce.^[96] Unfortunately, a photograph or likeness, on its own, typically does not qualify for trademark protection, unless it is consistently used on specific goods.^[97] Even where there is a likeness (or similar “mark”) used consistently as part of branding a distinct product, many trademark-infringement claims would be difficult to establish in this context, because trademark law does little to protect many aspects of a creator’s work.

Moreover, the Supreme Court has been wary about creating a sort of “mutant copyright” in cases that invoke the Lanham Act as a means to enforce a sort of “right of attribution,” which would potentially give creators the ability to control the use of their name in broader contexts.^[98] In this context, the Court has held that the relevant parts of the Lanham Act were not designed to “protect originality or creativity,”^[99] but are focused solely on “actions like trademark infringement that deceive consumers and impair a producer’s goodwill.”^[100]

In many ways, there is a parallel here to the trademark cases involving keyword bidding in online ads. At a high level, search engines and other digital-advertising services do not generally infringe trademark when they allow businesses to purchase ads triggered by a user’s search for competitor trademarks (i.e., rivals’ business names).^[101] But in some contexts, this can be infringing—e.g., where the use of trademarked terms in combination with advertising text can mislead consumers about the origin of a good or service.^[102]

Thus, the harm, when it arises, would not be in a user asking an AI system to generate something “in the style of” a known creator, but when that user subsequently seeks to release a new AI-generated work and falsely claims it originated from the creator, or leaves the matter ambiguous and misleading to consumers.

Alternative remedies for creators could be found in the “right of publicity” laws in various states. A state-level right of publicity “is not merely a legal right of the ‘celebrity,’ but is a right inherent to everyone to control the commercial use of identity and persona and recover in court damages and the commercial value of an unpermitted taking.”^[103] Such rights are recognized under state common law and statutes, which vary considerably in scope across jurisdictions—frequently as part of other privacy statutes.^[104] For example, some states only protect an individual’s name, likeness, or voice, while others also cover distinctive appearances, gestures, and mannerisms.^[105] The protections afforded for right-of-publicity claims vary significantly based on the state where the unauthorized use occurs or the individual is domiciled.^[106] This creates challenges for the application of uniform nationwide protection of creators’ interests in the various aspects that such laws protect.

In recent hearings before the U.S. Senate Judiciary Subcommittee on Intellectual Property, several witnesses advocated creating a federal version of the right of publicity.^[107] The Copyright Office has also previously opined that it may be desirable for Congress to enact some form of a “right of publicity” law.^[108] If Congress chose to enact a federal “right of privacy” statute, several key issues would need to be addressed regarding the scope of protection, effect on state laws, constitutional authority, and First Amendment limitations.

Congress would have to delineate the contours of the federal right of publicity, including the aspects of identity covered and the types of uses prohibited. A broad right of privacy could protect names, images, likenesses, voices, gestures, distinctive appearances, and biographical information from any unauthorized commercial use. Or Congress could take a narrower approach focused only on particular identity attributes, like name and likeness. Congress would also need to determine whether a federal right-of-publicity statute preempts state right-of-publicity laws or sets a floor that would allow state protections to exceed the federal standards.

4.              Bargaining for the use of likenesses

A federal right of publicity could present an interesting way out of the current dispute between rightsholders and AI producers. Most of the foregoing comment attempts to pull apart different pieces of potential infringement actions, but such actions are only necessary, obviously, if a mutually beneficial agreement cannot be struck between creators and AI producers. The main issue at hand is that, given the vast amount of content necessary to train an AI system, it could be financially impractical for even the largest AI firms to license all the necessary content. Even if the comments above are correct, and fair use is not available, it could very well be the case that AI producers will not license very much content, possibly relying on public-domain material, and choosing to license only a very small selection.

Something like a “right of publicity,” or an equivalent agreement between creators and AI producers, could provide alternative licensing and monetization strategies that encourage cooperation between the parties. If creators had the opportunity to opt into the use of their likeness (or the relevant equivalent for the sort of AI system in question), the creators could generate revenue when the AI system actually uses the results of processing their content. Thus, the producers would not need to license content that contributes an unknown and possibly de minimis value to their systems, and would only need to pay for individual instances of use.

Indeed, in this respect, we are already beginning to see some experimentation with business models. The licensing of celebrity likenesses for Meta’s new AI chatbots highlights an emerging opportunity for creators to monetize their brand through contractual agreements that grant usage rights to tech companies that commercialize conversational AI.[109] As this technology matures, there will be more opportunities for collaborations between AI producers—who are eager to leverage reputable and recognizable personalities—and celebrities or influencers seeking new income streams.

As noted, much of the opportunity for creators and AI producers to reach these agreements will depend on how rights are assigned.[110] It may be the case that a “right of publicity” is not necessary to make this sort of bargaining happen, as creators could—at least theoretically—pursue litigation on a state-by-state basis. This disparate-litigation strategy could deter many creators, however, and it could also be the case that a single federal standard outlining a minimal property right in “publicity” could help to facilitate bargaining.

Conclusion

The advent of generative AI systems presents complex new public-policy challenges centered on the intersection of technology and copyright law. As the Copyright Office’s inquiry recognizes, there are open questions around the legal status of AI-training data, the attribution of AI outputs, and infringement liability, which all require thoughtful analysis.

Ultimately, maintaining incentives for human creativity, while also allowing AI systems to flourish, will require compromise and cooperation between stakeholders. Rather than an outright ban on the unauthorized use of copyrighted works for training data, a licensing market that enables access to a large corpora could emerge. Rightsholders may need to accept changes to how they typically license content. In exchange, AI producers will have to consider how they can share the benefit of their use of protected works with creators.

Copyright law retains flexibility to adapt to new technologies, as past reforms reacting to photography, sound recordings, software, and the internet all demonstrate. With careful balancing of interests, appropriate limitations, and respect for constitutional bounds, copyright can continue to promote the progress of science and the useful arts even in the age of artificial intelligence. This inquiry marks a constructive starting point, although ongoing reassessment will likely be needed as generative AI capabilities continue to advance rapidly.

[1] Artificial Intelligence and Copyright, Notice of Inquiry and Request for Comments, U.S. Copyright Office, Library of Congress (Aug. 30, 2023) [hereinafter “NOI”].

^[2] Tim Sweeney (@TimSweeneyEpic), Twitter (Jan. 15, 2023, 3:35 AM), https://twitter.com/timsweeneyepic/status/1614541807064608768?s=46&t=0MH_nl5w4PJJl46J2ZT0Dw.

^[3] Pulitzer Prize Winner and Other Authors Accuse OpenAI of Misusing Their Writing, Competition Policy International (Sep. 11, 2023), https://www.pymnts.com/cpi_posts/pulitzer-prize-winner-and-other-authors-accuse-openai-of-misusing-their-writing; Getty Images Statement, Getty Images (Jan. 17, 2023), https://newsroom.gettyimages.com/en/getty-images/getty-images-statement.

[4] See, e.g., Anton Oleinik, What Are Neural Networks Not Good At? On Artificial Creativity, 6 Big Data & Society (2019), available at https://journals.sagepub.com/doi/full/10.1177/2053951719839433#bibr75-2053951719839433.

[5] William M. Landes & Richard A. Posner, An Economic Analysis of Copyright Law, 18 J. Legal Stud. 325 (1989).

[6] Id. at 332.

[7] Id. at 326.

[8] Id.

[9] See infra, notes 102-103 and accompanying text.

^[10] See generally R.H. Coase, The Problem of Social Cost, 3 J. L. & Econ. 1, 2 (1960).

^[11] Richard Posner, Economic Analysis of Law (Aspen 5^th ed 1998) 65, 79.

^[12] Coase, supra note 9, at 27.

^[13] Id.

^[14] Id. at 27.

^[15] Id. at 42-43.

^[16] U.S. Copyright Office, Library of Congress, supra note 1, at 14.

^[17] For more detailed discussion of GANs and Stable Diffusion see Ian Spektor, From DALL E to Stable Diffusion: How Do Text-to-image Generation Models Work?, Tryo Labs Blog (Aug. 31, 2022), https://tryolabs.com/blog/2022/08/31/from-dalle-to-stable-diffusion.

^[18] Id.

^[19] Id.

^[20] Id.

^[21] Id.

^[22] Id.

^[23] Jay Alammar, The Illustrated Stable Diffusion, Blog (Oct. 4, 2022), https://jalammar.github.io/illustrated-stable-diffusion.

^[24] Indeed, there is evidence that some models may be trained in a way that they “memorize” their training set, to at least some extent. See, e.g., Kent K. Chang, Mackenzie Cramer, Sandeep Soni, & David Bamman, Speak, Memory: An Archaeology of Books Known to ChatGPT/GPT-4, arXiv Preprint (Oct. 20, 2023), https://arxiv.org/abs/2305.00118; OpenAI LP, Comment Regarding Request for Comments on Intellectual Property Protection for Artificial Intelligence Innovation, Before the USPTO, Dep’t of Com. (2019), available at https://www.uspto.gov/sites/default/files/documents/OpenAI_RFC-84-FR-58141.pdf.

^[25] OpenAI, LP, Comment Regarding Request for Comments on Intellectual Property Protection for Artificial Intelligence, id. (emphasis added).

^[26] 17 U.S.C. § 107.

^[27] See, e.g., Blake Brittain, Meta Tells Court AI Software Does Not Violate Author Copyrights, Reuters (Sep. 19, 2023), https://www.reuters.com/legal/litigation/meta-tells-court-ai-software-does-not-violate-author-copyrights-2023-09-19; Avram Piltch, Google Wants AI Scraping to be ‘Fair Use.’ Will That Fly in Court?, Tom’s Hardware (Aug. 11, 2023), https://www.tomshardware.com/news/google-ai-scraping-as-fair-use.

^[28] 17 U.S.C. § 106.

^[29] Register of Copyrights, DMCA Section 104 Report (U.S. Copyright Office, Aug. 2001), at 108-22, available at https://www.copyright.gov/reports/studies/dmca/sec-104-report-vol-1.pdf.

^[30] Id. at 122-23.

^[31] Id. at 112 (emphasis added).

[32] Id. at 129–30.

[33] 17 U.S.C. § 107.

[34] Id.; see also Campbell v. Acuff-Rose Music Inc., 510 U.S. 569 (1994).

[35] Critically, a fair use analysis is a multi-factor test, and even within the first factor, it’s not a mandatory requirement that a use be “transformative.” It is entirely possible that a court balancing all of the factors could indeed find that training AI systems is fair use, even if it does not hold that such uses are “transformative.”

[36] Campbell, supra note 22, at 591.

[37] Authors Guild v. Google, Inc., 804 F.3d 202, 214 (2d Cir. 2015).

[38] OpenAI submission, supra note 13, at 5.

[39] Id. at 915.

[40] Id.

[41] Id.

^[42] Id. at 933-34.

^[43] Id. at 923. (emphasis added)

^[44] Id.

[45] Id. at 924.

[46] Kelly v. Arriba Soft Corp., 336 F.3d 811 (9th Cir. 2002).

[47] Id.

[48] Id. at 818.

[49] Id.

[50] Id. at 819 (“Arriba’s use of the images serves a different function than Kelly’s use—improving access to information on the internet versus artistic expression.”).

[51] The “public benefit” aspect of copyright law is reflected in the fair-use provision, 17 U.S.C. § 107. In Campbell v. Acuff-Rose Music, Inc., 510 U.S. 569, 579 (1994), the Supreme Court highlighted the “social benefit” that a use may provide, depending on the first of the statute’s four fair-use factors, the “the purpose and character of the use.”

[52] Supra note 46, at 820.

[53] Perfect 10 Inc. v. Amazon.com Inc., 487 F.3d 701 (9^th Cir., 2007)

[54] Id. at 721 (“Although an image may have been created originally to serve an entertainment, aesthetic, or informative function, a search engine transforms the image into a pointer directing a user to a source of information.”).

[55] Id. at 721.

[56] Id. at 723 (emphasis added).

[57] Id. (emphasis added).

[58] Id.

[59] Supra note 37, at 218.

[60] Id. at 215-16.

[61] Id. at 214. See also id. (“The more the appropriator is using the copied material for new, transformative purposes, the more it serves copyright’s goal of enriching public knowledge and the less likely it is that the appropriation will serve as a substitute for the original or its plausible derivatives, shrinking the protected market opportunities of the copyrighted work.”).

[62] Id. at 218.

[63] Perfect 10, 487 F.3d at 721-22 (citing Kelly, 336 F.3d at 818-19). See also Campbell, 510 U.S. at 579 (“The central purpose of this investigation is to see, in Justice Story’s words, whether the new work merely ‘supersede[s] the objects’ of the original creation, or instead adds something new, with a further purpose or different character….”) (citations omitted).

[64] See supra, notes 9-14 and accompanying text.

[65] See, e.g., the development of the compulsory “mechanical royalty,” now embodied in 17 U.S.C. § 115, that was adopted in the early 20^th century as a way to make it possible for the manufacturers of player pianos to distribute sheet music playable by their instruments.

[66] See supra notes 9-14 and accompanying text.

^[67] U.S. Copyright Office, Library of Congress, supra note 1, at 15.

[68] See infra, notes at 102-103 and accompanying text.

^[69] U.S. Copyright Office, Library of Congress, supra note 1, at 15.

^[70] See, e.g., Copyright Free Music, Premium Beat By Shutterstock, https://www.premiumbeat.com/royalty-free/licensed-music; Royalty-free stock footage at your fingertips, Adobe Stock, https://stock.adobe.com/video.

^[71] U.S. Copyright Office, Library of Congress, supra note 1, at 19.

^[72] Id.

^[73] U.S. Const. art. I, § 8, cl. 8.

[74] See Ajay Agrawal, Joshua S. Gans, & Avi Goldfarb, Exploring the Impact of Artificial Intelligence: Prediction Versus Judgment, 47 Info. Econ. & Pol’y 1, 1 (2019) (“We term this process of understanding payoffs, ‘judgment’. At the moment, it is uniquely human as no machine can form those payoffs.”).

^[75] Joshua Gans, Can AI works get copyright protection? (Redux), Joshua Gans’ Newsletter (Sept. 7, 2023), https://joshuagans.substack.com/p/can-ai-works-get-copyright-protection.

^[76] See Nicholas Carlini, et al., Extracting Training Data from Diffusion Models, Cornell Univ. (Jan. 30, 2023), available at https://arxiv.org/abs/2301.13188.

^[77] See Chang, Cramer, Soni, & Bamman, supra note 24; see also Matthew Sag, Copyright Safety for Generative AI, Working Paper (May 4, 2023), available at https://ssrn.com/abstract=4438593.; Andrés Guadamuz, A Scanner Darkly: Copyright Liability and Exceptions in Artificial Intelligence Inputs and Outputs, 25-27 (Mar. 1, 2023), available at https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4371204.

^[78] Laureyssens v. Idea Grp. Inc., 964 F.2d 131, 140 (2d Cir. 1992), as amended (June 24, 1992).

^[79] Id. at 139.

^[80] Harney v. Sony Pictures Television Inc., 704 F.3d 173, 178 (1st Cir. 2013). This assumes, for argument’s sake, that a given model is not “memorizing,” as noted above.

^[81] Id. at 178-79.

^[82] Id.

[83] U.S. Copyright Office, Library of Congress, supra note 1, at 25.

^[84] Notably, the state of mind of the user would be irrelevant from the point of view of whether an infringement occurs. All that is required is that a plaintiff owns a valid copyright, and that the defendant infringed it. 17 U.S.C. 106. There are cases where the state of mind of the defendant will matter, however. For one, willful or recklessly indifferent infringement by a plaintiff will open the door for higher statutory damages. See, e.g., Island Software & Computer Serv., Inc. v. Microsoft Corp., 413 F.3d 257, 263 (2d Cir. 2005). For another, a case of criminal copyright infringement will require that a defendant have acted “willfully.” 17 U.S.C. § 506(a)(1) (2023), 18 U.S.C. § 2319 (2023).

^[85] Id. at 550.

[86] Legally speaking, it would be incoherent to suggest that there can be secondary liability without primary liability. The way that AI systems work, however, could prompt Congress to modify the law in order to account for the identified situation.

^[87] See, e.g., Metro-Goldwyn-Mayer Studios Inc. v. Grokster Ltd., 380 F.3d 1154, 1160 (9th Cir. 2004), vacated and remanded, 545 U.S. 913, 125 S. Ct. 2764, 162 L. Ed. 2d 781 (2005).

^[88] See BMG Rts. Mgmt. (US) LLC v. Cox Commc’ns Inc., 881 F.3d 293, 306 (4th Cir. 2018); Sony Corp. of Am. v. Universal City Studios Inc., 464 U.S. 417, 442 (1984).

^[89] A&M Recs. Inc. v. Napster Inc., 239 F.3d 1004, 1022 (9th Cir. 2001), as amended (Apr. 3, 2001), aff’d sub nom. A&M Recs. Inc. v. Napster Inc., 284 F.3d 1091 (9th Cir. 2002), and aff’d sub nom. A&M Recs. Inc. v. Napster Inc., 284 F.3d 1091 (9th Cir. 2002).

^[90] See, e.g., Content Filtering, Microsoft Ignite, available at https://learn.microsoft.com/en-us/azure/ai-services/openai/concepts/content-filter (last visited Oct. 27, 2023).

^[91] Note that, if an AI producer can demonstrate that they used no protected works in the training phase, there may in fact be no liability for infringement at all. If a protected work is never made available to the AI system, even an output very similar to that protected work might not be “substantially similar” in a legal sense.

^[92] Copyright Review Board, Second Request for Reconsideration for Refusal to Register Théâtre D’opéra Spatial (SR # 1-11743923581; Correspondence ID: 1-5T5320R), U.S. Copyright Office (Sep. 5, 2023), available at https://fingfx.thomsonreuters.com/gfx/legaldocs/byprrqkqxpe/AI%20COPYRIGHT%20REGISTRATION%20decision.pdf.

^[93] Am. Broad. Companies Inc. v. Aereo Inc., 573 U.S. 431, 453 (2014). (Thomas J, dissenting).

^[94] Id. at 451.

[95] U.S. Copyright Office, Library of Congress, supra note 1, at 21.

^[96] See 5 U.S.C. § 1051 et seq. at § 1127.

^[97] See, e.g., ETW Corp. v. Jireh Pub. Inc., 332 F.3d 915, 923 (6th Cir. 2003).

^[98] Dastar Corp. v. Twentieth Century Fox Film Corp., 539 U.S. 23, 34 (2003).

^[99] Id. at 37.

^[100] Id. at 32.

^[101] See, e.g., Multi Time Mach. Inc. v. Amazon.com Inc., 804 F.3d 930, 938 (9^th Cir. 2015); EarthCam Inc. v. OxBlue Corp., 49 F. Supp. 3d 1210, 1241 (N.D. Ga. 2014); Coll. Network Inc. v. Moore Educ. Publishers Inc., 378 F. App’x 403, 414 (5th Cir. 2010).

^[102] Digby Adler Grp. LLC v. Image Rent a Car Inc., 79 F. Supp. 3d 1095, 1102 (N.D. Cal. 2015).

^[103] J. Thomas McCarthy, The Rights of Publicity and Privacy § 1:3. Introduction—Definition and History of the Right of Publicity—Simple Definition of the Right of Publicity, 1 Rights of Publicity and Privacy § 1:3 (2d ed).

^[104] See id. at § 6:3.

^[105] Compare Ind. Code § 32-36-1-7 (covering name, voice, signature, photograph, image, likeness, distinctive appearance, gesture, or mannerism), with Ky. Rev. Stat. Ann. § 391.170 (limited to name and likeness for “public figures”).

^[106] See Restatement (Third) of Unfair Competition § 46 (1995).

^[107] See, e.g., Jeff Harleston, Artificial Intelligence and Intellectual Property – Part II: Copyright, U.S. Senate Comm. on the Judiciary Subcomm. on Intellectual Property (Jul.12, 2023), available at https://www.judiciary.senate.gov/imo/media/doc/2023-07-12_pm_-_testimony_-_harleston1.pdf; Karla Ortiz, “AI and Copyright”, U.S. Senate Comm. on the Judiciary Subcomm. on Intellectual Property (Jul. 7, 2023), available at https://www.judiciary.senate.gov/imo/media/doc/2023-07-12_pm_-_testimony_-_ortiz.pdf; Matthew Sag, “Artificial Intelligence and Intellectual Property – Part II: Copyright and Artificial Intelligence”, U.S. Senate Comm. on the Judiciary Subcomm. on Intellectual Property (Jul. 12, 2023), available at https://www.judiciary.senate.gov/imo/media/doc/2023-07-12_pm_-_testimony_-_sag.pdf.

^[108] Authors, Attribution, and Integrity: Examining Moral Rights in the United States, U.S. Copyright Office (Apr. 2019) at 117-119, https://www.copyright.gov/policy/moralrights/full-report.pdf.

[109] Benj Edwards, Meta Launches Consumer AI Chatbots with Celebrity Avatars in its Social Apps, ArsTechnica (Sep. 28, 2023), https://arstechnica.com/information-technology/2023/09/meta-launches-consumer-ai-chatbots-with-celebrity-avatars-in-its-social-apps; Max Chafkin, Meta’s New AI Buddies Aren’t Great Conversationalists, Bloomberg (Oct. 17, 2023), https://www.bloomberg.com/news/newsletters/2023-10-17/meta-s-celebrity-ai-chatbots-on-facebook-instagram-are-surreal.

[110] See supra, notes 8-14 and accompanying text.

ICLE Director of Innovation Policy Kristian Stout appeared as a guest on Minnesota Public Radio’s Marketplace in a segment on the Federal Communications Commission’s decision to reinstate so-called “net neutrality” for broadband providers.

But Kristian Stout, director of innovation policy at the International Center for Law and Economics, argues that we don’t need net neutrality as much as we once did because most of us are already online now. So how do we ensure access for every last American?

“You don’t do that by upending or frustrating the investment incentives that have made this work really well for 90 to 95% of the country. What you do is try to figure out targeted solutions,” Stout said.

Audio of the full segment is embedded below.

We write to offer our thoughts in response to the National Telecommunications and Information Administration’s (NTIA) recently submitted comments to the Federal Communications Commission (FCC) in response to the notice of proposed rulemaking (NPRM) in this matter. Among its recommendations, the NTIA proposes that the FCC:^[1]

1. Adopt a disparate-impact standard to define “digital discrimination of access;” and
2. Subject a “broad range” of service characteristics to digital-discrimination rules, including pricing, promotional conditions, terms of service, and quality of service.

We urge the FCC to reject both of these recommendations. As we note in a recent International Center for Law & Economics (ICLE) issue brief, a disparate-impact approach would likely run afoul of the U.S. Supreme Court’s tests for when such an approach is appropriate.^[2] In addition, the NTIA’s recommendation to use digital-discrimination rules to regulate prices and quality of service would subject broadband-internet providers to rate regulation that the FCC has historically said it eschews.

A.      Section 60506 Does Not Support Disparate Impact Analysis

Section 60506 of the Infrastructure Investment and Jobs Act (IIJA) requires the FCC to adopt final rules facilitating equal access to broadband internet:

[T]he Commission shall adopt final rules to facilitate equal access to broadband internet access service, taking into account the issues of technical and economic feasibility presented by that objective, including … preventing digital discrimination of access based on income level, race, ethnicity, color, religion, or national origin; and… identifying necessary steps for the Commissions to take to eliminate discrimination …^[3]

The U.S. Supreme Court has established tests governing when it is appropriate to conduct an effects-based “disparate impact” analysis in the context of discrimination law. First, the presence of language like “otherwise make unavailable” is critical to construing a statute as demanding an effects-based analysis.^[4] Such phrases, the Court found, “refer[] to the consequences of an action rather than the actor’s intent.” Second, the structure of a statute’s language matters:

The relevant statutory phrases … play an identical role in the structure common to all three statutes: Located at the end of lengthy sentences that begin with prohibitions on disparate treatment, they serve as catchall phrases looking to consequences, not intent. And all [of these] statutes use the word “otherwise” to introduce the results-oriented phrase. “Otherwise” means “in a different way or manner,” thus signaling a shift in emphasis from an actor’s intent to the consequences of his actions.^[5]

Thus, as Section 60506 was drafted without “results-oriented language” and instead frames the prohibition against digital discrimination as “based on income level, race, ethnicity, color, religion, or national origin,” this would put the rule squarely within the realm of prohibitions on intentional discrimination. That is, to be discriminatory, the conduct in question must have been made intentionally because of the protected characteristic. Mere statistical correlation between outcomes and protected characteristics is insufficient to demonstrate discrimination under Section 60506.

B.      Broadband Rate Regulation Is at Odds With Longstanding FCC Policy

The FCC has for years been explicit about its apprehension to impose direct rate regulation on broadband-internet providers. Obama-era FCC Chair Tom Wheeler promised to forebear from rate regulation under the 2015 Open Internet Order (OIO), declaring “we are not trying to regulate rates.”^[6] Last month, in a speech announcing the FCC’s proposal to regulate broadband internet under Title II of the Communications Act, Chair Jessica Rosenworcel was emphatic: “They say this is a stalking horse for rate regulation. Nope. No how, no way.”^[7]

And yet, the NTIA recommends precisely that: a stalking horse for rate regulation under the guise of preventing digital discrimination: “Without addressing pricing as a possible source of discrimination, the Commission will be hard pressed to meet its statutory mandate to prevent digital discrimination of access.”^[8] Indeed, in addition to prices, the NTIA recommends addressing promotions, as well as terms of service and quality of service—both of which are inextricably intertwined with pricing. Section 60506 is explicit that the FCC must “tak[e] into account the issues of technical and economic feasibility” in addressing digital discrimination. The NTIA, however, appears to suggest ignoring that mandate.^[9]

C.      The Problem With De Facto Rate Regulation

In a competitive market, prices allow for the successful coordination of supply and demand, and the market price reflects both consumer demand and the costs of production. Of course, for those on the demand side of the equation, the price of a good or service is a cost to them, and they would prefer falling prices to rising prices. For suppliers, the price represents the revenue from selling the good or service, and they would prefer rising prices to falling prices.

Due to this inherent tension, there is a natural inclination on the part of both consumers and producers to seek government intervention in the competitive process to either halt or slow price changes. The most obvious way the government can intervene is through rate regulation, such as price controls. It is well known and widely accepted that price controls can make both consumers and sellers worse off.^[10] Consequently, policymakers often pitch policies to control prices under another name (e.g., “second-generation rent relief” instead of “rent control”) or introduce policies that are not explicit price controls. These de facto rate regulations  (e.g., quality-of-service mandates)  have substantially the same effects as direct price controls.

For example, some agricultural products are subject to “marketing orders,” which are legal cartels than can dictate the price and quality of produce.^[11] Consider an apple market subject to a marketing order that specifies fresh apples must be of a certain shape and size, such that only large, round apples can be sold as fresh produce.

Consumers presumably prefer large apples to small apples and prefer round apples to misshapen apples. Thus, the order that only large, round apples can be sold as fresh has the effect of increasing/shifting the demand curve. Consumers would be willing to pay more for the seemingly better fruit, and they’d be willing to buy more. But the order also increases the cost to apple growers. They have to find a way to dispose of their smaller or misshapen apples, perhaps by making apple sauce or juicing the fruit. They also incur higher costs of managing their crop to produce more of the higher-quality fruit. This has the effect of decreasing/shifting the supply curve for fresh fruit. Growers will supply less fruit at a higher cost.

Combining the effects from both the shift in supply and the shift in demand shows that the marketing order unambiguously results in a higher price for apples. What is not known, however, is whether more or fewer apples are sold. That will depend on the elasticities of demand and supply. Because the order results in a higher price, however, it has created a de facto price floor without explicitly setting one. Consumers are not aware that they are paying a higher price, because they do not know what type of fruit would otherwise be available, and at what price, absent the quality restrictions.

Similarly, broadband quality-of-service mandates simultaneously increase demand while increasing costs. Were all other things held constant, the result would be a higher price for broadband. What is not known, however, is whether more or fewer households will subscribe to broadband. That will depend on the elasticities of demand and supply. Because the mandate results in a higher price, however, it has created a de facto price floor without explicitly setting one. Consumers are not aware that they are paying a higher price because they do not know what quality of service would otherwise be available—and at what price—absent the quality-of-service mandate.

A recent ICLE issue brief explores in detail how these sorts of  terms-of-service and quality-of-service mandates often amount to de facto rate regulation.^[12] Sadly, we have seen many recent attempts—including by the NTIA itself—to introduce these sorts of de facto rate regulations:

• The NTIA’s notice of funding opportunity (NOFO) under the Broadband Equity, Access, and Deployment (BEAD) program requires each participating U.S. state or territory to include a “middle-class affordability plan to ensure that all consumers have access to affordable high-speed internet” (emphasis in original).^[13] The NOFO specifies a price (“affordable”); a quantity (“all middle-class households”); and imposes a quality mandate (“high-speed”).
• In its third and fourth funding rounds, the U.S. Department of Agriculture’s ReConnect Loan and Grant Program included provision of a “low-cost option” as a point criteria in award decisions. It also included a requirement that projects must provide broadband access at speeds of at least 100/100 Mbps (e., 100 Mbps symmetrical speed).
• The FCC’s 2015 Open Internet Order outright prohibited “paid prioritization”—that is, seeking payments for network utilization from edge providers like Google, Facebook, and Netflix—while casting suspicion on other pricing schemes under its Internet Conduct Standard.

Moreover, as we note in our Income Conundrum issue brief, the evaluation of digital-discrimination claims based on income level can yield highly complicated analyses due to income’s correlation with a host of factors, both protected (e.g., race and national origin) and unprotected (e.g., home-computer ownership).  Adoption of Section 60506 rules that do not recognize this “income conundrum” will invite costly and time-consuming disparate-impact litigation that alleges digital discrimination, both where no such discrimination exists and where it is excused by economic-feasibility considerations. Even worse, regulatory interventions on price, quality of service, and terms of service would extend this damage further by creating de facto utility regulation on providers that completely distorts investment incentives.

D.     Conclusion

The FCC must be cautious when promulgating rules under Section 60506. In particular, the commission should adopt an intent-based discriminatory-treatment standard, rather than one that opens the doors to disparate-impact claims. And FCC rules should articulate a presumption of nondiscrimination, in which allegations of digital discrimination must be demonstrated, rather than a presumption of discrimination that must be rebutted for each deployment, service, and pricing decision.

If the commission has good evidence of intentional discrimination in the deployment of broadband, it has a role to play in preventing it. But without strong, compelling evidence of intentional discrimination, the FCC will run the risk of a constitutional challenge to its rules and waste scarce resources chasing bogeymen.

Above all, the FCC should resist calls to engage in rate regulation, either through direct intervention on broadband prices or through interventions on quality of service or terms of service. The longstanding policy to avoid rate regulation has been an important factor leading to increased broadband deployment in the United States.

^[1] Ex Parte Comments of the National Telecommunications and Information Administration, In the Matter of Implementing the Infrastructure Investment and Jobs Act: Prevention and Elimination of Digital Discrimination, GN Docket No. 22-69 (Oct. 6, 2023), available at  https://www.ntia.gov/sites/default/files/publications/ntia_digital_discrimination_ex_parte_comment_10.6.23.pdf at 3 and 8.

^[2] Eric Fruits & Kristian Stout, The Income Conundrum: Intent and Effects Analysis of Digital Discrimination, Int’l. Ctr. for L. & Econ. (Nov. 14, 2022), available at https://laweconcenter.org/wp-content/uploads/2022/11/The-Income-Conundrum-Intent-and-Effects-Analysis-of-Digital-Discrimination.pdf; see also Eric Fruits, Kristian Stout, & Ben Sperry, ICLE Reply Comments on Prevention and Elimination of Digital Discrimination, Notice of Proposed Rulemaking, In the Matter of Implementing the Infrastructure, Investment, and Jobs Act: Prevention and Elimination of Digital Discrimination, No. 22-69, at Part III (Apr. 20, 2023), https://laweconcenter.org/resources/icle-reply-comments-on-prevention-and-elimination-of-digital-discrimination.

^[3] 47 U.S.C. § 1754 (emphasis added).

^[4] Texas Dep’t of Hous. & Cmty. Affs., 576 U.S. at 534.

^[5] Id. at 534-535 [emphasis added].

^[6] Tom Wheeler, Hearing on FCC Reauthorization: Oversight of the Commission, U.S. House Energy and Commerce Subcommittee on Communications and Technology, (Mar. 19, 2015), https://www.govinfo.gov/content/pkg/CHRG-114hhrg95817/html/CHRG-114hhrg95817.htm.

^[7] FCC Chair Rosenworcel on Reinstating Net Neutrality Rules, C-Span (Sep. 26, 2023), https://www.c-span.org/video/?530731-1/fcc-chair-rosenworcel-reinstating-net-neutrality-rule.

^[8] Supra note 1, at 10.

^[9] Id. (“Congress set out a specific list of demographic groups protected by this statute, including … racial and ethnic minorities (who as previously noted are disproportionately likely to live in environments where networks are costlier to maintain, among other challenges).”) [emphasis added].

^[10] See, e.g., N. Gregory Mankiw, Principles of Microeconomics, 4th ed., Thomson South-Western (2007); Paul Krugman & Robin Wells, Economics, 6th ed., MacMillan (2021); Steven A. Greenlaw & David Shapiro, Principles of Microeconomics 2nd ed., OpenStax (2017).

^[11] See Darren Filson, Edward Keen, Eric Fruits, & Thomas Borcherding, Market Power and Cartel Formation: Theory and an Empirical Test, 44 J. L. & ECON. 465 (2001).

^[12] Eric Fruits & Geoffrey A. Manne, Quack Attack: De Facto Rate Regulation in Telecommunications, Int’l. Ctr. for L. & Econ. (Mar. 30, 2023), available at https://laweconcenter.org/wp-content/uploads/2023/03/De-Facto-Rate-Reg-Final-1.pdf.

^[13] Notice of Funding Opportunity, Broadband Equity, Access, and Deployment Program, NTIA-BEAD-2022, NTIA (May 2022), available at https://broadbandusa.ntia.doc.gov/sites/default/files/2022-05/BEAD%20NOFO.pdf (note that the IIJA itself did not include this requirement, but it was added by NTIA as part of the NOFO process; thus, it is unclear the extent to which this represents a valid requirement by NTIA under the BEAD program).