Showing 9 of 126 Publications in Copyright

Systemic Risk and Copyright in the EU AI Act

TOTM The European Parliament’s approval last week of the AI Act marked a significant milestone in the regulation of artificial intelligence. While the law’s final text . . .

The European Parliament’s approval last week of the AI Act marked a significant milestone in the regulation of artificial intelligence. While the law’s final text is less alarming than what was initially proposed, it nonetheless still includes some ambiguities that could be exploited by regulators in ways that would hinder innovation in the EU. 

Among the key features emerging from the legislation are its introduction of “general purpose AI” (GPAI) as a regulatory category and the ways that these GPAI might interact with copyright rules. Moving forward in what is rapidly becoming a global market for generative-AI services, it also bears reflecting on how the AI Act’s copyright provisions contrast with current U.S. copyright law. 

Read the full piece here.

Continue reading
Intellectual Property & Licensing

Utility, Copyright, and Fair Use after Warhol

Scholarship Abstract This paper is a reaction to AWE v. Goldsmith (Warhol), which found that Warhol’s adaptation of a photograph of Prince, taken by photographer Lynn . . .

Abstract

This paper is a reaction to AWE v. Goldsmith (Warhol), which found that Warhol’s adaptation of a photograph of Prince, taken by photographer Lynn Goldsmith, is not protected from copyright liability by the fair use defense. The Warhol dissent accuses the majority of being overly concerned with the commercial character of Warhol’s use, while the dissent emphasizes the artistically transformative quality of Warhol’s adaptation. These different approaches provide strong evidence that the theory of fair use remains unclear to the Court. There is a need for a simple positive theory of thefair use doctrine. That need was largely met by Gordon’s article in 1982. I aim to develop the economic theory of fair use further. especially in light of case law since 1982. A theory of fair use is at the same time a theory of the scope of copyright. I clarify the economic basis for jair use, taking advantage of basic concepts in welfare economics. As a general matter, the optimal scope of copyright minimizes the sum of dynamic (having to do with incentives over time) and static (having to do with allocation at a given time) welfare costs. One proposition advanced is that the concepts of economic complementarity, substitutability, and preference correlation provide crucial analytical tools in resolving fair use disputes. This proposition may seem narrow, but it stands the approach taken in the cases on its head. I explain how the approach urged here works by applying it to several cases, including Warhol and Google v. Oracle.

Continue reading
Intellectual Property & Licensing

Jonah Gelbach on Free PACER

Presentations & Interviews ICLE Academic Affiliate Jonah Gelbach was a guest on Berkeley Law’s Voices Carry podcast to discuss how aggregating federal court data can help researchers tease . . .

ICLE Academic Affiliate Jonah Gelbach was a guest on Berkeley Law’s Voices Carry podcast to discuss how aggregating federal court data can help researchers tease out critical trends, as well as efforts to push the federal judiciary to drop the paywall on the Public Access to Court Electronic Records database. The full episode is embedded below.

Continue reading
Intellectual Property & Licensing

A Competition Law & Economics Analysis of Sherlocking

ICLE White Paper Abstract Sherlocking refers to an online platform’s use of nonpublic third-party business data to improve its own business decisions—for instance, by mimicking the successful products . . .

Abstract

Sherlocking refers to an online platform’s use of nonpublic third-party business data to improve its own business decisions—for instance, by mimicking the successful products and services of edge providers. Such a strategy emerges as a form of self-preferencing and, as with other theories about preferential access to data, it has been targeted by some policymakers and competition authorities due to the perceived competitive risks originating from the dual role played by hybrid platforms (acting as both referees governing their platforms, and players competing with the business they host). This paper investigates the competitive implications of sherlocking, maintaining that an outright ban is unjustified. First, the paper shows that, by aiming to ensure platform neutrality, such a prohibition would cover scenarios (i.e., the use of nonpublic third-party business data to calibrate business decisions in general, rather than to adopt a pure copycat strategy) that should be analyzed separately. Indeed, in these scenarios, sherlocking may affect different forms of competition (inter-platform v. intra-platform competition). Second, the paper argues that, in either case, the practice’s anticompetitive effects are questionable and that the ban is fundamentally driven by a bias against hybrid and vertically integrated players.

I. Introduction

The dual role some large digital platforms play (as both intermediary and trader) has gained prominence among the economic arguments used to justify the recent wave of regulation hitting digital markets around the world. Many policymakers have expressed concern about potential conflicts of interest among companies that have adopted this hybrid model and that also control important gateways for business users. In other words, the argument goes, some online firms act not only as regulators who set their platforms’ rules and as referees who enforce those rules, but also as market players who compete with their business users. This raises the fear that large platforms could reserve preferential treatment for their own services and products, to the detriment of downstream rivals and consumers. That, in turn, has led to calls for platform-neutrality rules.

Toward this aim, essentially all of the legislative initiatives undertaken around the world in recent years to enhance competition in digital markets have included anti-discrimination provisions that target various forms of self-preferencing. Self-preferencing, it has been said, serves as the symbol of the current competition-policy zeitgeist in digital markets.[1] Indeed, this conduct is considered functional to leveraging strategies that would grant gatekeepers the chance to entrench their power in core markets and extend it into associated markets.[2]

Against this background, so-called “sherlocking” has emerged as one form of self-preferencing. The term was coined roughly 20 years ago, after Apple updated its own app Sherlock (a search tool on its desktop-operating system) to mimic a third-party application called Watson, which was created by Karelia Software to complement the Apple tool’s earlier version.[3] According to critics of self-preferencing generally and sherlocking in particular, biased intermediation and related conflicts of interest allow gatekeepers to exploit their preferential access to business users’ data to compete against them by replicating successful products and services. The implied assumption is that this strategy is relevant to competition policy, even where no potential intellectual-property rights (IPRs) are infringed and no slavish imitation sanctionable under unfair-competition laws is detected. Indeed, under such theories, sherlocking would already be prevented by the enforcement of these rules.

To tackle perceived misuse of gatekeepers’ market position, the European Union’s Digital Markets Act (DMA) introduced a ban on sherlocking.[4] Similar concerns have also motivated requests for intervention in the United States,[5] Australia,[6] and Japan.[7] In seeking to address at least two different theories of gatekeepers’ alleged conflicts of interest, these proposed bans on exploiting access to business users’ data are not necessarily limited to the risk of product imitation, but may include any business decision whatsoever that a platform may make while relying on that data.

In parallel with the regulatory initiatives, the conduct at-issue has also been investigated in some antitrust proceedings, which appear to seek the very same twofold goal. In particular, in November 2020, the European Commission sent a statement of objections to Amazon that argued the company had infringed antitrust rules through the systematic use of nonpublic business data from independent retailers who sell on the Amazon online marketplace in order to benefit Amazon’s own retail business, which directly competes with those retailers.[8] A similar investigation was opened by the UK Competition and Markets Authority (CMA) in July 2022.[9]

Further, as part of the investigation opened into Apple’s App Store rule requiring developers to use Apple’s in-app purchase mechanism to distribute paid apps and/or paid digital content, the European Commission also showed interest in evaluating whether Apple’s conduct might disintermediate competing developers from relevant customer data, while Apple obtained valuable data about those activities and its competitors’ offers.[10] The European Commission and UK CMA likewise launched an investigation into Facebook Marketplace, with accusations that Meta used data gathered from advertisers in order to compete with them in markets where the company is active, such as classified ads.[11]

There are two primary reasons these antitrust proceedings are relevant. First, many of the prohibitions envisaged in regulatory interventions (e.g., DMA) clearly took inspiration from the antitrust investigations, thus making it important to explore the insights that competition authorities may provide to support an outright ban. Second, given that regulatory intervention will be implemented alongside competition rules (especially in Europe) rather than displace them,[12] sherlocking can be assessed at both the EU and national level against dominant players that are not eligible for “gatekeeper” designation under the DMA. For those non-gatekeeper firms, the practice may still be investigated by antitrust authorities and assessed before courts, aside from the DMA’s per se prohibition. And, of course, investigations and assessments of sherlocking could also be made even in those jurisdictions where there isn’t an outright ban.

The former sis well-illustrated by the German legislature’s decision to empower its national competition authority with a new tool to tackle abusive practices that are similar and functionally equivalent to the DMA.[13] Indeed, as of January 2021, the Bundeskartellamt may identify positions of particular market relevance (undertakings of “paramount significance for competition across markets”) and assess their possible anticompetitive effects on competition in those areas of digital ecosystems in which individual companies may have a gatekeeper function. Both the initiative’s aims and its list of practices are similar to the DMA. They are distinguished primarily by the fact that the German list is exhaustive, and the practices at-issue are not prohibited per se, but are subject to a reversal of the burden of proof, allowing firms to provide objective justifications. For the sake of this analysis, within the German list, one provision prohibits designated undertakings from “demanding terms and conditions that permit … processing data relevant for competition received from other undertakings for purposes other than those necessary for the provision of its own services to these undertakings without giving these undertakings sufficient choice as to whether, how and for what purpose such data are processed.”[14]

Unfortunately, none of the above-mentioned EU antitrust proceedings have concluded with a final decision that addresses the merits of sherlocking. This precludes evaluating whether the practice would have survived before the courts. Regarding the Apple investigation, the European Commission dropped the case over App Store rules and issued a new statement of objections that no longer mentions sherlocking.[15] Further, the European Commission and the UK CMA accepted the commitments offered by Amazon to close those investigations.[16] The CMA likewise accepted the commitments offered by Meta.[17]

Those outcomes can be explained by the DMA’s recent entry into force. Indeed, because of the need to comply with the new regulation, players designated as gatekeepers likely have lost interest in challenging antitrust investigations that target the very same conduct prohibited by the DMA.[18] After all, given that the DMA does not allow any efficiency defense against the listed prohibitions, even a successful appeal against an antitrust decision would be a pyrrhic victory. From the opposite perspective, the same applies to the European Commission, which may decide to save time, costs, and risks by dropping an ongoing case against a company designated as a gatekeeper under the DMA, knowing that the conduct under investigation will be prohibited in any case.

Nonetheless, despite the lack of any final decision on sherlocking, these antitrust assessments remain relevant. As already mentioned, the DMA does not displace competition law and, in any case, dominant platforms not designated as gatekeepers under the DMA still may face antitrust investigations over sherlocking. This applies even more for jurisdictions, such as the United States, that are evaluating DMA-like legislative initiatives (e.g., the American Innovation and Choice Online Act, or “AICOA”).

Against this background, drawing on recent EU cases, this paper questions the alleged anticompetitive implications of sherlocking, as well as claims that the practice fails to comply with existing antitrust rules.

First, the paper illustrates that prohibitions on the use of nonpublic third-party business data would cover two different theories that should be analyzed separately. Whereas a broader case involves all the business decisions adopted by a dominant platform because of such preferential access (e.g., the launch of new products or services, the development or cessation of existing products or services, the calibration of pricing and management systems), a more specific case deals solely with the adoption of a copycat strategy. By conflating these theories in support of a blanket ban that condemns any use of nonpublic third-party business data, EU antitrust authorities are fundamentally motivated by the same policy goal pursued by the DMA—i.e., to impose a neutrality regime on large online platforms. The competitive implications differ significantly, however, as adopting copycat strategies may only affect intra-brand competition, while using said data to improve other business decisions could also affect inter-platform competition.

Second, the paper shows that, in both of these scenarios, the welfare effects of sherlocking are unclear. Notably, exploiting certain data to better understand the market could help a platform to develop new products and services, to improve existing products and services, or more generally to be more competitive with respect to both business users and other platforms. As such outcomes would benefit consumers in terms of price and quality, any competitive advantage achieved by the hybrid platform could be considered unlawful only if it is not achieved on the merits. In a similar vein, if sherlocking is used by a hybrid platform to deliver replicas of its business users’ products and services, that would likely provide short-term procompetitive effects benefitting consumers with more choice and lower prices. In this case, the only competitive harm that would justify an antitrust intervention resides in (uncertain) negative long-term effects on innovation.

As a result, in any case, an outright ban of sherlocking, such as is enshrined in the DMA, is economically unsound since it would clearly harm consumers.

The paper is structured as follows. Section II describes the recent antitrust investigations of sherlocking, illustrating the various scenarios that might include the use of third-party business data. Section III investigates whether sherlocking may be considered outside the scope of competition on the merits for bringing competitive advantages to platforms solely because of their hybrid business model. Section IV analyzes sherlocking as a copycat strategy by investigating the ambiguous welfare effects of copying in digital markets and providing an antitrust assessment of the practice at issue. Section V concludes.

II. Antitrust Proceedings on Sherlocking: Platform Neutrality and Copycat Competition

Policymakers’ interest in sherlocking is part of a larger debate over potentially unfair strategies that large online platforms may deploy because of their dual role as an unavoidable trading partner for business users and a rival in complementary markets.

In this scenario, as summarized in Table 1, the DMA outlaws sherlocking, establishing that to “prevent gatekeepers from unfairly benefitting from their dual role,”[19] they are restrained from using, in competition with business users, “any data that is not publicly available that is generated or provided by those business users in the context of their use of the relevant core platform services or of the services provided together with, or in support of, the relevant core platform services, including data generated or provided by the customers of those business users.”[20] Recital 46 further clarifies that the “obligation should apply to the gatekeeper as a whole, including but not limited to its business unit that competes with the business users of a core platform service.”

A similar provision was included in the American Innovation and Choice Online Act (AICOA), which was considered, but not ultimately adopted, in the 117th U.S. Congress. AICOA, however, would limit the scope of the ban to the offer of products or services that would compete with those offered by business users.[21] Concerns about copycat strategies were also reported in the U.S. House of Representatives’ investigation of the state of competition in digital markets as supporting the request for structural-separation remedies and line-of-business restrictions to eliminate conflicts of interest where a dominant intermediary enters markets that place it in competition with dependent businesses.[22] Interestingly, however, in the recent complaint filed by the U.S. Federal Trade Commission (FTC) and 17 state attorneys general against Amazon that accuses the company of having deployed an interconnected strategy to block off every major avenue of competition (including price, product selection, quality, and innovation), there is no mention of sherlocking among the numerous unfair practices under investigation.[23]

Evaluating regulatory-reform proposals for digital markets, the Australian Competition and Consumer Commission (ACCC) also highlighted the risk of sherlocking, arguing that it could have an adverse effect on competition, notably on rivals’ ability to compete, when digital platforms exercise their strong market position to utilize nonpublic data to free ride on the innovation efforts of their rivals.[24] Therefore, the ACCC suggested adopting service-specific codes to address self-preferencing by, for instance, imposing data-separation requirements to restrain dominant app-store providers from using commercially sensitive data collected from the app-review process to develop their own apps.[25]

Finally, on a comparative note, it is also useful to mention the proposals advanced by the Japanese Fair Trade Commission (JFTC) in its recent market-study report on mobile ecosystems.[26] In order to ensure equal footing among competitors, the JFTC specified that its suggestion to prevent Google and Apple from using nonpublic data generated by other developers’ apps aims at pursuing two purposes. Such a ban would, indeed, concern not only use of the data for the purpose of developing competing apps, products, and services, but also its use for developing their own apps, products, and services.

TABLE 1: Legislative Initiatives and Proposals to Ban Sherlocking

As previously anticipated, sherlocking recently emerged as an antitrust offense in three investigations launched by the European Commission and the UK CMA.

In the first case, Amazon’s alleged reliance on marketplace sellers’ nonpublic business data has been claimed to distort fair competition on its platform and prevent effective competition. In its preliminary findings, the Commission argued that Amazon takes advantage of its hybrid business model, leveraging its access to nonpublic third-party sellers’ data (e.g., the number of ordered and shipped units of products; sellers’ revenues on the marketplace; the number of visits to sellers’ offers; data relating to shipping, to sellers’ past performance, and to other consumer claims on products, including the activated guarantees) to adjust its retail offers and strategic business decisions to the detriment of third-party sellers, which are direct competitors on the marketplace.[27] In particular, the Commission was concerned that Amazon uses such data for its decision to start and end sales of a product, for its pricing system, for its inventory-planning and management system, and to identify third-party sellers that Amazon’s vendor-recruitment teams should approach to invite them to become direct suppliers to Amazon Retail. To address the data-use concern, Amazon committed not to use nonpublic data relating to, or derived from, independent sellers’ activities on its marketplace for its retail business and not to use such data for the purposes of selling branded goods, as well as its private-label products.[28]

A parallel investigation ended with similar commitments in the UK.[29] According to the UK CMA, Amazon’s access to and use of nonpublic seller data could result in a competitive advantage for Amazon Retail arising from its operation of the marketplace, rather than from competition on the merits, and may lead to relevant adverse effects on competition. Notably, it was alleged this could result in a reduction in the scale and competitiveness of third-party sellers on the Amazon Marketplace; a reduction in the number and range of product offers from third-party sellers on the Amazon Marketplace; and/or less choice for consumers, due to them being offered lower quality goods and/or paying higher prices than would otherwise be the case.

It is also worth mentioning that, by determining that Amazon is an undertaking of paramount significance for competition across markets, the Bundeskartellamt emphasized the competitive advantage deriving from Amazon’s access to nonpublic data, such as Glance Views, sales figures, sale quantities, cost components of products, and reorder status.[30] Among other things, with particular regard to Amazon’s hybrid role, the Bundeskartellamt noted that the preferential access to competitively sensitive data “opens up the possibility for Amazon to optimize its own-brand assortment.”[31]

A second investigation involved Apple and its App Store rule.[32] According to the European Commission, the mandatory use of Apple’s own proprietary in-app purchase system (IAP) would, among other things, grant Apple full control over the relationship its competitors have with customers, thus disintermediating those competitors from customer data and allowing Apple to obtain valuable data about the activities and offers of its competitors.

Finally, Meta faced antitrust proceedings in both the EU and the UK.[33] The focus was on Facebook Marketplace—i.e., an online classified-ads service that allows users to advertise goods for sale. According to the European Commission and the CMA, Meta unilaterally imposes unfair trading conditions on competing online-classified ads services that advertise on Facebook or Instagram. These terms and conditions, which authorize Meta to use ads-related data derived from competitors for the benefit of Facebook Marketplace, are considered unjustified, as they impose an unnecessary burden on competitors and only benefit Facebook Marketplace. The suspicion is that Meta has used advertising data from Facebook Marketplace competitors for the strategic planning, product development, and launch of Facebook Marketplace, as well as for Marketplace’s operation and improvement.

Overall, these investigations share many features. The concerns about third-party business-data use, as well as about other forms of self-preferencing, revolve around the competitive advantages that accrue to a dominant platform because of its dual role. Such advantages are considered unfair, as they are not the result of the merits of a player, but derived purely and simply from its role as an important gateway to reach end users. Moreover, this access to valuable business data is not reciprocal. The feared risk is the marginalization of business users competing with gatekeepers on the gatekeepers’ platforms and, hence, the alleged harm to competition is the foreclosure of rivals in complementary markets (horizontal foreclosure).

The focus of these investigations was well-illustrated by the European Commission’s decision on Amazon’s practice.[34] The Commission’s concern was about the “data delta” that Amazon may exploit, namely the additional data related to third-party sellers’ listings and transactions that are not available to, and cannot be replicated by, the third-party sellers themselves, but are available to and used by Amazon Retail for its own retail operations.[35] Contrary to Amazon Retail—which, according to Commission’s allegations, would have full access to and would use such individual, real-time data of all its third-party sellers to calibrate its own retail decisions—sellers would have access only to their own individual listings and sales data. As a result, the Commission came to the (preliminary) conclusion that real-time access to and use of such volume, variety, and granularity of non-publicly available data from its retail competitors generates a significant competitive advantage for Amazon Retail in each of the different decisional processes that drive its retail operations.[36]

On a closer look, however, while antitrust authorities seem to target the use of nonpublic third-party business data as a single theory of harm, their allegations cover two different scenarios along the lines of what has already been examined with reference to the international legislative initiatives and proposals. Indeed, the Facebook Marketplace case does not involve an allegation of copying, as Meta is accused of gathering data from its business users to launch and improve its ads service, instead of reselling goods and services.

FIGURE 1: Sherlocking in Digital Markets

As illustrated above in Figure 1, while the claim in the latter scenario is that the preferential data use would help dominant players calibrate business decisions in general, the former scenario instead involves the use of such data for a pure copycat strategy of an entire product or service, or some of its specific features.

In both scenarios the aim of the investigations is to ensure platform neutrality. Accordingly, as shown by the accepted commitments, the envisaged solution for antitrust authorities is to impose  data-separation requirements to restrain dominant platforms from using third-party commercially sensitive data. Putting aside that these investigations concluded with commitments from the firms, however, their chances of success before a court differ significantly depending on whether they challenge a product-imitation strategy, or any business decision adopted because of the “data delta.”

A. Sherlocking and Unconventional Theories of Harm for Digital Markets

Before analyzing how existing competition-law rules could be applied to the various scenarios involving the use of third-party business data, it is worth providing a brief overview of the framework in which the assessment of sherlocking is conducted. As competition in the digital economy is increasingly a competition among ecosystems,[37] a lively debate has emerged on the capacity of traditional antitrust analysis to adequately capture the peculiar features of digital markets. Indeed, the combination of strong economies of scale and scope; indirect network effects; data advantages and synergies across markets; and portfolio effects all facilitate ecosystem development all contribute to making digital markets highly concentrated, prone to tipping, and not easily contestable.[38] As a consequence, it’s been suggested that addressing these distinctive features of digital markets requires an overhaul of the antitrust regime.

Such discussions require the antitrust toolkit and theories of harm to illustrate whether and how a particular practice, agreement, or merger is anticompetitive. Notably, at issue is whether traditional antitrust theories of harm are fit for purpose or whether novel theories of harm should be developed in response to the emerging digital ecosystems. The latter requires looking at the competitive impact of expanding, protecting, or strengthening an ecosystem’s position, and particularly whether such expansion serves to exploit a network of capabilities and to control access to key inputs and components.[39]

A significant portion of recent discussions around developing novel theories of harm to better address the characteristics of digital-business models and markets has been devoted to the topic of merger control—in part a result of the impressive number of acquisitions observed in recent years.[40] In particular, the focus has been on analyzing conglomerate mergers that involve acquiring a complementary or unrelated asset, which have traditionally been assumed to raise less-significant competition concerns.

In this regard, an ecosystem-based theory seems to have guided the Bundeskartellamt in its assessment of Meta’s acquisition of Kustomer[41] and by the CMA in Microsoft/Activision.[42] A more recent example is the European Commission’s decision to prohibit the proposed Booking/eTraveli merger, where the Commission explicitly noted that the transaction would have allowed Booking to expand its travel-services ecosystem.[43] The Commission’s concerns were related primarily to the so-called “envelopment” strategy, in which a prominent platform within a specific market broadens its range of services into other markets where there is a significant overlap of customer groups already served by the platform.[44]

Against this background, putative self-preferencing harms represent one of the European Commission’s primary (albeit contentious)[45] attempts to develop new theories of harm built on conglomerate platforms’ ability to bundle services or use data from one market segment to inform product development in another.[46] Originally formulated in the Google Shopping decision,[47] the theory of harm of (leveraging through) self-preferencing has subsequently inspired the DMA, which targets different forms of preferential treatment, including sherlocking.

In particular, it is asserting that platform may use self-preferencing to adopt a leveraging strategy with a twofold anticompetitive effect—that is, excluding or impeding rivals from competing with the platform (defensive leveraging) and extending the platform’s market power into associated markets (offensive leveraging). These goals can be pursued because of the unique role that some large digital platforms play. That is, they not only enjoy strategic market status by controlling ecosystems of integrated complementary products and services, which are crucial gateways for business users to reach end users, but they also perform a dual role as both a critical intermediary and a player active in complementors’ markets. Therefore, conflicts of interests may provide incentives for large vertically integrated platforms to favor their own products and services over those of their competitors.[48]

The Google Shopping theory of harm, while not yet validated by the Court of Justice of the European Union (CJEU),[49] has also found its way into merger analysis, as demonstrated by the European Commission’s recent assessment of iRobot/Amazon.[50] In its statement of objections, the Commission argued that the proposed acquisition of iRobot may give Amazon the ability and incentive to foreclose iRobot’s rivals by engaging in several foreclosing strategies to prevent them from selling robot vacuum cleaners (RVCs) on Amazon’s online marketplace and/or at degrading such rivals’ access to that marketplace. In particular, the Commission found that Amazon could deploy such self-preferencing strategies as delisting rival RVCs; reducing rival RVCs’ visibility in both organic and paid results displayed in Amazon’s marketplace; limiting access to certain widgets or commercially attractive labels; and/or raising the costs of iRobot’s rivals to advertise and sell their RVCs on Amazon’s marketplace.[51]

Sherlocking belongs to this framework of analysis and can be considered a form of self-preferencing, specifically because of the lack of reciprocity in accessing sensitive data.[52] Indeed, while gatekeeper platforms have access to relevant nonpublic third-party business data as a result of their role as unavoidable trading partners, they leverage this information exclusively, without sharing it with third-party sellers, thus further exacerbating an already uneven playing field.[53]

III. Sherlocking for Competitive Advantage: Hybrid Business Model, Neutrality Regimes, and Competition on the Merits

Insofar as prohibitions of sherlocking center on the competitive advantages that platforms enjoy because of their dual role—thereby allowing some players to better calibrate their business decisions due to their preferential access to business users’ data—it should be noted that competition law does not impose a general duty to ensure a level playing field.[54] Further, a competitive advantage does not, in itself, amount to anticompetitive foreclosure under antitrust rules. Rather, foreclosure must not only be proved (in terms of actual or potential effects) but also assessed against potential benefits for consumers in terms of price, quality, and choice of new goods and services.[55]

Indeed, not every exclusionary effect is necessarily detrimental to competition.[56] Competition on the merits may, by definition, lead to the departure from the market or the marginalization of competitors that are less efficient and therefore less attractive to consumers from the point of view of, among other things, price, choice, quality or innovation.[57] Automatically classifying any conduct with exclusionary effects were as anticompetitive could well become a means to protect less-capable, less-efficient undertakings and would in no way protect more meritorious undertakings—thereby potentially hindering a market’s competitiveness.[58]

As recently clarified by the CJEU regarding the meaning of “competition on the merits,” any practice that, in its implementation, holds no economic interest for a dominant undertaking except that of eliminating competitors must be regarded as outside the scope of competition on the merits.[59] Referring to the cases of margin squeezes and essential facilities, the CJEU added that the same applies to practices that a hypothetical equally efficient competitor is unable to adopt because that practice relies on using resources or means inherent to the holding of such a dominant position.[60]

Therefore, while antitrust cases on sherlocking set out to ensure a level playing field and platform neutrality, and therefore center on the competitive advantages that a platform enjoys because of its dual role, mere implementing a hybrid business model does not automatically put such practices outside the scope of competition on the merits. The only exception, according to the interpretation provided in Bronner, is the presence of an essential facility—i.e., an input whose access should be considered indispensable, as there are no technical, legal, or economic obstacles capable of making it impossible, or even unreasonably difficult, to duplicate it.[61]

As a result, unless it is proved that the hybrid platform is an essential facility, sherlocking and other forms of self-preferencing cannot be considered prima facie outside the scope of competition on the merits, or otherwise unlawful. Rather, any assessment of sherlocking demands the demonstration of anticompetitive effects, which in turn requires finding an impact on efficient firms’ ability and incentive to compete. In the scenario at-issue, for instance, the access to certain data may allow a platform to deliver new products or services; to improve existing products or services; or more generally to compete more efficiently not only with respect to the platform’s business users, but also against other platforms. Such an increase in both intra-platform and inter-platform competition would benefit consumers in terms of lower prices, better quality, and a wider choice of new or improved goods and services—i.e., competition on the merits.[62]

In Facebook Marketplace, the European Commission and UK CMA challenged the terms and conditions governing the provision of display-advertising and business-tool services to which Meta required its business customers to sign up.[63] In their view, Meta abused its dominant position by imposing unfair trading conditions on its advertising customers, which authorized Meta to use ads-related data derived from the latter in a way that could afford Meta a competitive advantage on Facebook Marketplace that would not have arisen from competition on the merits. Notably, antitrust authorities argued that Meta’s terms and conditions were unjustified, disproportionate, and unnecessary to provide online display-advertising services on Meta’s platforms.

Therefore, rather than directly questioning the platform’s dual role or hybrid business model, the European Commission and UK CMA decided to rely on traditional case law which considers unfair those clauses that are unjustifiably unrelated to the purpose of the contract, unnecessarily limit the parties’ freedom, are disproportionate, or are unilaterally imposed or seriously opaque.[64] This demonstrates that, outside the harm theory of the unfairness of terms and conditions, a hybrid platform’s use of nonpublic third-party business data to improve its own business decisions is generally consistent with antitrust provisions. Hence, an outright ban would be unjustified.

IV. Sherlocking to Mimic Business Users’ Products or Services

The second, and more intriguing, sherlocking scenario is illustrated by the Amazon Marketplace investigations and regards the original meaning of sherlocking—i.e., where a data advantage is used by a hybrid platform to mimic its business users’ products or services.

Where sherlocking charges assert that the practice allows some platforms to use business users’ data to compete against them by replicating their products or services, it should not be overlooked that the welfare effects of such a copying strategy are ambiguous. While the practice could benefit consumers in the short term by lowering prices and increasing choice, it may discourage innovation over the longer term if third parties anticipate being copied whenever they deliver successful products or services. Therefore, the success of an antitrust investigation essentially relies on demonstrating a harm to innovation that would induce business users to leave the market or stop developing their products and services. In other words, antitrust authorities should be able to demonstrate that, by allowing dominant platforms to free ride on their business guests’ innovation efforts, sherlocking would negatively affect rivals’ ability to compete.

A. The Welfare Effects of Copying

The tradeoff between the short- and long-term welfare effects of copying has traditionally been analyzed in the context of the benefits and costs generated by intellectual-property protection.[65] In particular, the economic literature investigating the optimal life of patents[66] and copyrights[67] focuses on the efficient balance between dynamic benefits associated with innovation and the static costs of monopoly power granted by IPRs.

More recently, product imitation has instead been investigated in the different scenario of digital markets, where dominant platforms adopting a hybrid business model may use third-party sellers’ market data to design and promote their own products over their rivals’ offerings. Indeed, some studies report that large online platforms may attempt to protect their market position by creating “kill zones” around themselves—i.e., by acquiring, copying, or eliminating their rivals.[68] In such a novel setting, the welfare effects of copying are assessed regardless of the presence and the potential enforcement of IPRs, but within a strategy aimed at excluding rivals by exploiting the dual role of both umpire and player to get preferential access to sensitive data and free ride on their innovative efforts.[69]

Even in this context, however, a challenging tradeoff should be considered. Indeed, while in the short term, consumers may benefit from the platform’s imitation strategy in terms of lower prices and higher quality, they may be harmed in the longer term if third parties are discouraged from delivering new products and services. As a result, while there is empirical evidence on hybrid platforms successfully entering into third parties’ adjacent market segments, [70] the extant academic literature finds the welfare implications of such moves to be ambiguous.

A first strand of literature attempts to estimate the welfare impact of the hybrid business model. Notably, Andre Hagiu, Tat-How Teh, and Julian Wright elaborated a model to address the potential implications of an outright ban on platforms’ dual mode, finding that such a structural remedy may harm consumer surplus and welfare even where the platform would otherwise engage in product imitation and self-preferencing.[71] According to the authors, banning the dual mode does not restore the third-party seller’s innovation incentives or the effective price competition between products, which are the putative harms caused by imitation and self-preferencing. Therefore, the authors’ evaluation was that interventions specifically targeting product imitation and self-preferencing were preferable.

Germa?n Gutie?rrez suggested that banning the dual model would generate hardly any benefits for consumers, showing that, in the Amazon case, interventions that eliminate either the Prime program or product variety are likely to decrease welfare.[72]

Further, analyzing Amazon’s business model, Federico Etro found that the platform and consumers’ incentives are correctly aligned, and that Amazon’s business model of hosting sellers and charging commissions prevents the company from gaining through systematic self?preferencing for its private-label and first-party products.[73] In the same vein, on looking at its business model and monetization strategy, Patrick Andreoli-Versbach and Joshua Gans argued that Amazon does not have an obvious incentive to self-preference.[74] Indeed, Amazon’s profitability data show that, on average, the company’s operating margin is higher on third-party sales than on first-party retail sales.

Looking at how modeling details may yield different results with regard to the benefits and harms of the hybrid business model, Simon Anderson and O?zlem Bedre-Defoile maintain that the platform’s choice to sell its own products benefits consumers by lowering prices when a monopoly platform hosts competitive fringe sellers, regardless of the platform’s position as a gatekeeper, whether sellers have an alternate channel to reach consumers, or whether alternate channels are perfect or imperfect substitutes for the platform channel.[75] On the other hand, the authors argued that platform product entry might harm consumers when a big seller with market power sells on its own channel and also on the platform. Indeed, in that case, the platform setting a seller fee before the big seller prices its differentiated products introduces double markups on the big seller’s platform-channel price and leaves some revenue to the big seller.

Studying whether Amazon engages in self-preferencing on its marketplace by favoring its own brands in search results, Chiara Farronato, Andrey Fradkin, and Alexander MacKay demonstrate empirically that Amazon brands remain about 30% cheaper and have 68% more reviews than other similar products.[76] The authors acknowledge, however, that their findings do not imply that consumers are hurt by Amazon brands’ position in search results.

Another strand of literature specifically tackles the welfare effects of sherlocking. In particular, Erik Madsen and Nikhil Vellodi developed a theoretical framework to demonstrate that a ban on insider imitation can either stifle or stimulate innovation, depending on the nature of innovation.[77] Specifically, the ban could stimulate innovation for experimental product categories, while reducing innovation in incremental product markets, since the former feature products with a large chance of superstar demand and the latter generate mostly products with middling demand.

Federico Etro maintains that the tradeoffs at-issue are too complex to be solved with simple interventions, such as bans on dual mode, self-preferencing, or copycatting.[78] Indeed, it is difficult to conclude that Amazon entry is biased to expropriate third-party sellers or that bans on dual mode, self-preferencing, or copycatting would benefit consumers, because they either degrade services and product variety or induce higher prices or commissions.

Similar results are provided by Jay Pil Choi, Kyungmin Kim, and Arijit Mukherjee, who developed a tractable model of a platform-run marketplace where the platform charges a referral fee to the sellers for access to the marketplace, and may also subsequently launch its own private-label product by copying a seller.[79] The authors found that a policy to either ban hybrid mode or only prohibit information use for the launch of private-label products may produce negative welfare implications.

Further, Radostina Shopova argues that, when introducing a private label, the marketplace operator does not have incentive to distort competition and foreclose the outside seller, but does have an incentive to lower fees charged to the outside seller and to vertically differentiate its own product in order to protect the seller’s channel.[80] Even when the intermediary is able to perfectly mimic the quality of the outside seller and monopolize its product space, the intermediary prefers to differentiate its offer and chooses a lower quality for the private-label product. Accordingly, as the purpose of private labels is to offer a lower-quality version of products aimed at consumers with a lower willingness to pay, a marketplace operator does not have an incentive to distort competition in favor of its own product and foreclose the seller of the original higher-quality product.

In addition, according to Jean-Pierre Dubé, curbing development of private-label programs would harm consumers and Amazon’s practices amount to textbook retailing, as they follow an off-the-shelf approach to managing private-label products that is standard for many retail chains in the West.[81] As a result, singling out Amazon’s practices would set a double standard.

Interestingly, such findings about predictors and effects of Amazon’s entry in competition with third-party merchants on its own marketplace are confirmed by the only empirical study developed so far. In particular, analyzing the Home & Kitchen department of Germany’s version of Amazon Marketplace between 2016 and 2021, Gregory S. Crawford, Matteo Courthoud, Regina Seibel, and Simon Zuzek’s results suggest that Amazon’s entry strategy was more consistent with making Marketplace more attractive to consumers than expropriating third-party merchants.[82] Notably, the study showed that, comparing Amazon’s entry decisions with those of the largest third-party merchants, Amazon tends to enter low-growth and low-quality products, which is consistent with a strategy that seeks to make Marketplace more attractive by expanding variety, lessening third-party market power, and/or enhancing product availability. The authors therefore found that Amazon’s entry on Amazon Marketplace demonstrated no systematic adverse effects and caused a mild market expansion.

Massimo Motta and Sandro Shelegia explored interactions between copying and acquisitions, finding that the former (or the threat of copying) can modify the outcome of an acquisition negotiation.[83] According to their model, there could be both static and dynamic incentives for an incumbent to introduce a copycat version of a complementary product. The static rationale consists of lowering the price of the complementary product in order to capture more rents from it, while the dynamic incentive consists of harming a potential rival’s prospects of developing a substitute. The latter may, in turn, affect the direction the entrant takes toward innovation. Anticipating the incumbent’s copying strategy, the entrant may shift resources from improvements to compete with the incumbent’s primary product to developing complementary products.

Jingcun Cao, Avery Haviv, and Nan Li analyzed the opposite scenario—i.e., copycats that seek to mimic the design and user experience of incumbents’ successful products.[84] The authors find empirically that, on average, copycat apps do not have a significant effect on the demand for incumbent apps and that, as with traditional counterfeit products, they may generate a positive demand spillover toward authentic apps.

Massimo Motta also investigated the potential foreclosure effects of platforms adopting a copycat strategy committed to non-discriminatory terms of access for third parties (e.g., Apple App Store, Google Play, and Amazon Marketplace).[85] Notably, according to Motta, when a third-party seller is particularly successful and the platform is unable to raise fees and commissions paid by that seller, the platform may prefer to copy its product or service to extract more profits from users, rather than rely solely on third-party sales. The author acknowledged, however, that even though this practice may create an incentive for self-preferencing, it does not necessarily have anticompetitive effects. Indeed, the welfare effects of the copying strategy are a priori ambiguous.[86] While, on the one hand, the platform’s copying of a third-party product benefits consumers by increasing variety and competition among products, on the other hand, copying might be wasteful for society, in that it entails a fixed cost and may discourage innovation if rivals anticipate that they will be systematically copied whenever they have a successful product.[87] Therefore, introducing a copycat version of a product offered by a firm in an adjacent market might be procompetitive.

B. Antitrust Assessment: Competition, Innovation, and Double Standards

The economic literature has demonstrated that the rationale and welfare effects of sherlocking by hybrid platforms are definitively ambiguous. Against concerns about rivals’ foreclosure, some studies provide a different narrative, illustrating that such a strategy is more consistent with making the platform more attractive to consumers (by differentiating the quality and pricing of the offer) than expropriating business users.[88] Furthermore, copies, imitations, and replicas undoubtedly benefit consumers with more choice and lower prices.

Therefore, the only way to consider sherlocking anticompetitive is by demonstrating long-term deterrent effects on innovation (i.e., reducing rivals’ incentives to invest in new products and services) outweigh consumers’ short-term advantages.[89] Moreover, deterrent effects must not be merely hypothetical, as a finding of abuse cannot be based on a mere possibility of harm.[90] In any case, such complex tradeoffs are at odds with a blanket ban.[91]

Moreover, assessments of the potential impact of sherlocking on innovation cannot disregard the role of IPRs—which are, by definition, the main primary to promote innovation. From this perspective, intellectual-property protection is best characterized as another form of tradeoff. Indeed, the economic rationale of IPRs (in particular, of patents and copyrights) involves, among other things, a tradeoff between access and incentives—i.e., between short-term competitive restrictions and long-term innovative benefits.[92]

According to the traditional incentive-based theory of intellectual property, free riding would represent a dangerous threat that justifies the exclusive rights granted by intellectual-property protection. As a consequence, so long as copycat expropriation does not infringe IPRs, it should be presumed legitimate and procompetitive. Indeed, such free riding is more of an intellectual-property issue than a competitive concern.

In addition, to strike a fair balance between restricting competition and providing incentives to innovation, the exclusive rights granted by IPRs are not unlimited in terms of duration, nor in terms of lawful (although not authorized) uses of the protected subject matter. Under the doctrine of fair use, for instance, reverse engineering represents a legitimate way to obtain information about a firm’s product, even if the intended result is to produce a directly competing product that may steer customers away from the initial product and the patented invention.

Outside of reverse engineering, copying is legitimately exercised once IPRs expire, when copycat competitors can reproduce previously protected elements. As a result of the competitive pressure exerted by new rivals, holders of expired IPRs may react by seeking solutions designed to block or at least limit the circulation of rival products. They could, for example, request other IPRs to cover aspects or functionalities different from those previously protected. They could also bring (sometimes specious) legal action for infringement of the new IPR or for unfair competition by slavish imitation. For these reasons, there have been occasions where copycat competitors have received protection from antitrust authorities against sham litigation brought by IPR holders concerned about losing margins due to pricing pressure from copycats.[93]

Finally, within the longstanding debate on the intersection of intellectual-property protection and competition, EU antitrust authorities have traditionally been unsympathetic toward restrictions imposed by IPRs. The success of the essential-facility doctrine (EFD) is the most telling example of this attitude, as its application in the EU has been extended to IPRs. As a matter of fact, the EFD represents the main antitrust tool for overseeing intellectual property in the EU.[94]

After Microsoft, EU courts have substantially dismantled one of the “exceptional circumstances” previously elaborated in Magill and specifically introduced for cases involving IPRs, with the aim of safeguarding a balance between restrictions to access and incentives to innovate. Whereas the CJEU established in Magill that refusal to grant an IP license should be considered anticompetitive if it prevents the emergence of a new product for which there is potential consumer demand, in Microsoft, the General Court considered such a requirement met even when access to an IPR is necessary for rivals to merely develop improved products with added value.

Given this background, recent competition-policy concerns about sherlocking are surprising. To briefly recap, the practice at-issue increases competition in the short term, but may affect incentives to innovate in the long-term. With regard to the latter, however, the practice neither involves products protected by IPRs nor constitutes a slavish imitation that may be caught under unfair-competition laws.

The case of Amazon, which has received considerable media coverage, is illustrative of the relevance of IP protection. Amazon has been accused of cloning batteries, power strips, wool runner shoes, everyday sling bags, camera tripods, and furniture.[95] One may wonder what kind of innovation should be safeguarded in these cases against potential copies. Admittedly, such examples appear consistent with the findings of the already-illustrated empirical study conducted by Crawford et al. indicating that Amazon tends to enter low-quality products in order to expand variety on the Marketplace and to make it more attractive to consumers.

Nonetheless, if an IPR is involved, right holders are provided with proper means to protect their products against infringement. Indeed, one of the alleged targeted companies (Williams-Sonoma) did file a complaint for design and trademark infringement, claiming that Amazon had copied a chair (Orb Dining Chair) sold by its West Elm brand. According to Williams-Sonoma, the Upholstered Orb Office Chair—which Amazon began selling under its Rivet brand in 2018—was so similar that the ordinary observer would be confused by the imitation.[96] If, instead, the copycat strategy does not infringe any IPR, the potential impact on innovation might not be considered particularly worrisome—at least at first glance.

Further, neither the degree to which third-party business data is unavailable nor the degree to which they are relevant in facilitating copying are clear cut. For instance, in the case of Amazon, public product reviews supply a great deal of information[97] and, regardless of the fact that a third party is selling a product on the Marketplace, anyone can obtain an item for the purposes of reverse engineering.[98]

In addition, antitrust authorities are used to intervening against opportunistic behavior by IPR holders. European competition authorities, in particular, have never before seemed particularly responsive to the motives of inventors and creators versus the need to encourage maximum market openness.

It should also be noted that cloning is a common strategy in traditional markets (e.g., food products)[99] and has been the subject of longstanding controversies between high-end fashion brands and fast-fashion brands (e.g., Zara, H&M).[100] Furthermore, brick-and-mortar retailers also introduce private labels and use other brands’ sales records in deciding what to produce.[101]

So, what makes sherlocking so different and dangerous when deployed in digital markets as to push competition authorities to contradict themselves?[102]

The double standard against sherlocking reflects the same concern and pursues the same goal of the various other attempts to forbid any form of self-preferencing in digital markets. Namely, antitrust investigations of sherlocking are fundamentally driven by the bias against hybrid and vertically integrated players. The investigations rely on the assumption that conflicts of interest have anticompetitive implications and that, therefore, platform neutrality should be promoted to ensure the neutrality of the competitive process.[103] Accordingly, hostility toward sherlocking may involve both of the illustrated scenarios—i.e., the use of nonpublic third-party business data either in adopting any business decision, or just copycat strategies, in particular.

As a result, however, competition authorities end up challenging a specific business model, rather than the specific practice at-issue, which brings undisputed competitive benefits in terms of lower prices and wider consumer choice, and which should therefore be balanced against potential exclusionary risks. As the CJEU has pointed out, the concept of competition on the merits:

…covers, in principle, a competitive situation in which consumers benefit from lower prices, better quality and a wider choice of new or improved goods and services. Thus, … conduct which has the effect of broadening consumer choice by putting new goods on the market or by increasing the quantity or quality of the goods already on offer must, inter alia, be considered to come within the scope of competition on the merits.[104]

Further, in light of the “as-efficient competitor” principle, competition on the merits may lead to “the departure from the market, or the marginalization of, competitors that are less efficient and so less attractive to consumers from the point of view of, among other things, price, choice, quality or innovation.”[105]

It has been correctly noted that the “as-efficient competitor” principle is a reminder of what competition law is about and how it differs from regulation.[106] Competition law aims to protect a process, rather than engineering market structures to fulfill a particular vision of how an industry is to operate.[107] In other words, competition law does not target firms on the basis of size or status and does not infer harm from (market or bargaining) power or business model. Therefore, neither the dual role played by some large online platforms nor their preferential access to sensitive business data or their vertical integration, by themselves, create a competition problem. Competitive advantages deriving from size, status, power, or business model cannot be considered per se outside the scope of competition on the merits.

Some policymakers have sought to resolve these tensions in how competition law regards sherlocking by introducing or envisaging an outright ban. These initiatives and proposals have clearly been inspired by antitrust investigations, but they did so for the wrong reasons. Instead of taking stock of the challenging tradeoffs between short-term benefits and long-term risks that an antitrust assessment of sherlocking requires, they blamed competition law for not providing effective tools to achieve the policy goal of platform neutrality.[108] Therefore, the regulatory solution is merely functional to bypass the traditional burden of proof of antitrust analysis and achieve what competition-law enforcement cannot provide.

V. Conclusion

The bias against self-preferencing strikes again. Concerns about hybrid platforms’ potential conflicts of interest have led policymakers to seek prohibitions to curb different forms of self-preferencing, making the latter the symbol of the competition-policy zeitgeist in digital markets. Sherlocking shares this fate. Indeed, the DMA outlaws any use of business users’ nonpublic data and similar proposals have been advanced in the United States, Australia, and Japan. Further, like other forms of self-preferencing, such regulatory initiatives against sherlocking have been inspired by previous antitrust proceedings.

Drawing on these antitrust investigations, the present research shows the extent to which an outright ban on sherlocking is unjustified. Notably, the practice at-issue includes two different scenarios: the broad case in which a gatekeeper exploits its preferential access to business users’ data to better calibrate all of its business decisions and the narrow case in which such data is used to adopt a copycat strategy. In either scenario, the welfare effects and competitive implications of sherlocking are unclear.

Indeed, the use of certain data by a hybrid platform to improve business decisions generally should be classified as competition on the merits, and may yield an increase in both intra-platform (with respect to business users) and inter-platform (with respect to other platforms) competition. This would benefit consumers in terms of lower prices, better quality, and a wider choice of new or improved goods and services. In a similar vein, if sherlocking is used to deliver replicas of business users’ products or services, the anti-competitiveness of such a strategy may only result from a cumbersome tradeoff between short-term benefits (i.e., lower prices and wider choice) and negative long-term effects on innovation.

An implicit confirmation of the difficulties encountered in demonstrating the anti-competitiveness of sherlocking comes from the recent complaint issued by the FTC against Amazon.[109] Current FTC Chairwoman Lina Khan devoted a significant portion of her previous academic career to questioning Amazon’s practices (including the decision to introduce its own private labels inspired by third-party products)[110] and to supporting the adoption of structural-separation remedies to tackle platforms’ conflicts of interest that induce them to exploit their “systemic informational advantage (gleaned from competitors)” to thwart rivals and strengthen their own position by introducing replica products.[111] Despite these premises and although the FTC’s complaint targets numerous practices belonging to what has been described as an interconnected strategy to block off every major avenue of competition, however, sherlocking is surprisingly off the radar.

Regulatory initiatives to ban sherlocking in order to ensure platform neutrality with respect to business users and a level playing field among rivals would sacrifice undisputed procompetitive benefits on the altar of policy goals that competition rules are not meant to pursue. Sherlocking therefore appears to be a perfect case study of the side effects of unwarranted interventions in digital markets.

[1] Giuseppe Colangelo, Antitrust Unchained: The EU’s Case Against Self-Preferencing, 72 GRUR International 538 (2023).

[2] Jacques Cre?mer, Yves-Alexandre de Montjoye, & Heike Schweitzer, Competition Policy for the Digital Era (2019), 7, https://op.europa.eu/en/publication-detail/-/publication/21dc175c-7b76-11e9-9f05-01aa75ed71a1/language-en (all links last accessed 3 Jan. 2024); UK Digital Competition Expert Panel, Unlocking Digital Competition, (2019) 58, available at https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/785547/unlocking_digital_competition_furman_review_web.pdf.

[3] You’ve Been Sherlocked, The Economist (2012), https://www.economist.com/babbage/2012/07/13/youve-been-sherlocked.

[4] Regulation (EU) 2022/1925 on contestable and fair markets in the digital sector and amending Directives (EU) 2019/1937 and (EU) 2020/1828 (Digital Markets Act) (2022), OJ L 265/1, Article 6(2).

[5] U.S. S. 2992, American Innovation and Choice Online Act (AICOA) (2022), Section 3(a)(6), available at https://www.klobuchar.senate.gov/public/_cache/files/b/9/b90b9806-cecf-4796-89fb-561e5322531c/B1F51354E81BEFF3EB96956A7A5E1D6A.sil22713.pdf. See also U.S. House of Representatives, Subcommittee on Antitrust, Commercial, and Administrative Law, Investigation of Competition in Digital Markets, Majority Staff Reports and Recommendations (2020), 164, 362-364, 378, available at https://democrats-judiciary.house.gov/uploadedfiles/competition_in_digital_markets.pdf.

[6] Australian Competition and Consumer Commission, Digital Platform Services Inquiry Report on Regulatory Reform (2022), 125, https://www.accc.gov.au/about-us/publications/serial-publications/digital-platform-services-inquiry-2020-2025/digital-platform-services-inquiry-september-2022-interim-report-regulatory-reform.

[7] Japan Fair Trade Commission, Market Study Report on Mobile OS and Mobile App Distribution (2023), https://www.jftc.go.jp/en/pressreleases/yearly-2023/February/230209.html.

[8] European Commission, 10 Nov. 2020, Case AT.40462, Amazon Marketplace; see Press Release, Commission Sends Statement of Objections to Amazon for the Use of Non-Public Independent Seller Data and Opens Second Investigation into Its E-Commerce Business Practices, European Commission (2020), https://ec.europa.eu/commission/presscorner/detail/en/ip_20_2077.

[9] Press Release, CMA Investigates Amazon Over Suspected Anti-Competitive Practices, UK Competition and Markets Authority (2022), https://www.gov.uk/government/news/cma-investigates-amazon-over-suspected-anti-competitive-practices.

[10] European Commission, 16 Jun. 2020, Case AT.40716, Apple – App Store Practices.

[11] Press Release, Commission Sends Statement of Objections to Meta over Abusive Practices Benefiting Facebook Marketplace, European Commission (2022), https://ec.europa.eu/commission/presscorner/detail/en/ip_22_7728; Press Release, CMA Investigates Facebook’s Use of Ad Data, UK Competition and Markets Authority (2021), https://www.gov.uk/government/news/cma-investigates-facebook-s-use-of-ad-data.

[12] DMA, supra note 4, Recital 10 and Article 1(6).

[13] GWB Digitalization Act, 18 Jan. 2021, Section 19a. On risks of overlaps between the DMA and the competition law enforcement, see Giuseppe Colangelo, The European Digital Markets Act and Antitrust Enforcement: A Liaison Dangereuse, 47 European Law Review 597.

[14] GWB, supra note 13, Section 19a (2)(4)(b).

[15] Press Release, Commission Sends Statement of Objections to Apple Clarifying Concerns over App Store Rules for Music Streaming Providers, European Commission (2023), https://ec.europa.eu/commission/presscorner/detail/en/ip_23_1217.

[16] European Commission, 20 Dec. 2022, Case AT.40462; Press Release, Commission Accepts Commitments by Amazon Barring It from Using Marketplace Seller Data, and Ensuring Equal Access to Buy Box and Prime, European Commission (2022), https://ec.europa.eu/commission/presscorner/detail/en/ip_22_7777; UK Competition and Markets Authority, 3 Nov. 2023, Case No. 51184, https://www.gov.uk/cma-cases/investigation-into-amazons-marketplace.

[17] UK Competition and Markets Authority, 3 Nov. 2023, Case AT.51013, https://www.gov.uk/cma-cases/investigation-into-facebooks-use-of-data.

[18] See, e.g., Gil Tono & Lewis Crofts (2022), Amazon Data Commitments Match DMA Obligations, EU’s Vestager Say, mLex (2022), https://mlexmarketinsight.com/news/insight/amazon-data-commitments-match-dma-obligation-eu-s-vestager-says (reporting that Commissioner Vestager stated that Amazon’s data commitments definitively appear to match what would be asked within the DMA).

[19] DMA, supra note 4, Recital 46.

[20] Id., Article 6(2) (also stating that, for the purposes of the prohibition, non-publicly available data shall include any aggregated and non-aggregated data generated by business users that can be inferred from, or collected through, the commercial activities of business users or their customers, including click, search, view, and voice data, on the relevant core platform services or on services provided together with, or in support of, the relevant core platform services of the gatekeeper).

[21] AICOA, supra note 5.

[22] U.S. House of Representatives, supra note 5; see also Lina M. Khan, The Separation of Platforms and Commerce, 119 Columbia Law Review 973 (2019).

[23] U.S. Federal Trade Commission, et al. v. Amazon.com, Inc., Case No. 2:23-cv-01495 (W.D. Wash., 2023).

[24] Australian Competition and Consumer Commission, supra note 6, 125.

[25] Id., 124.

[26] Japan Fair Trade Commission, supra note 7, 144.

[27] European Commission, supra note 8. But see also Amazon, Supporting Sellers with Tools, Insights, and Data (2021), https://www.aboutamazon.eu/news/policy/supporting-sellers-with-tools-insights-and-data (claiming that the company is just using aggregate (rather than individual) data: “Just like our third-party sellers and other retailers across the world, Amazon also uses data to run our business. We use aggregated data about customers’ experience across the store to continuously improve it for everyone, such as by ensuring that the store has popular items in stock, customers are finding the products they want to purchase, or connecting customers to great new products through automated merchandising.”)

[28] European Commission, supra note 16.

[29] UK Competition and Markets Authority, supra notes 9 and 16.

[30] Bundeskartellamt, 5 Jul. 2022, Case B2-55/21, paras. 493, 504, and 518.

[31] Id., para. 536.

[32] European Commission, supra note 10.

[33] European Commission, supra note 11; UK Competition and Markets Authority, supra note 11.

[34] European Commission, supra note 16. In a similar vein, see also UK Competition and Markets Authority, supra note 16, paras. 4.2-4.7.

[35] European Commission, supra note 16, para. 111.

[36] Id., para. 123.

[37] Cre?mer, de Montjoye, & Schweitzer, supra note 2, 33-34.

[38] See, e.g., Marc Bourreau, Some Economics of Digital Ecosystems, OECD Hearing on Competition Economics of Digital Ecosystems (2020), https://www.oecd.org/daf/competition/competition-economics-of-digital-ecosystems.htm; Amelia Fletcher, Digital Competition Policy: Are Ecosystems Different?, OECD Hearing on Competition Economics of Digital Ecosystems (2020).

[39] See, e.g., Cristina Caffarra, Matthew Elliott, & Andrea Galeotti, ‘Ecosystem’ Theories of Harm in Digital Mergers: New Insights from Network Economics, VoxEU (2023), https://cepr.org/voxeu/columns/ecosystem-theories-harm-digital-mergers-new-insights-network-economics-part-1 (arguing that, in merger control, the implementation of an ecosystem theory of harm would require assessing how a conglomerate acquisition can change the network of capabilities (e.g., proprietary software, brand, customer-base, data) in order to evaluate how easily competitors can obtain alternative assets to those being acquired); for a different view, see Geoffrey A. Manne & Dirk Auer, Antitrust Dystopia and Antitrust Nostalgia: Alarmist Theories of Harm in Digital Markets and Their Origins, 28 George Mason Law Review 1281(2021).

[40] See, e.g., Viktoria H.S.E. Robertson, Digital merger control: adapting theories of harm, (forthcoming) European Competition Journal; Caffarra, Elliott, & Galeotti, supra note 39; OECD, Theories of Harm for Digital Mergers (2023), available at www.oecd.org/daf/competition/theories-of-harm-for-digital-mergers-2023.pdf; Bundeskartellamt, Merger Control in the Digital Age – Challenges and Development Perspectives (2022), available at https://www.bundeskartellamt.de/SharedDocs/Publikation/EN/Diskussions_Hintergrundpapiere/2022/Working_Group_on_Competition_Law_2022.pdf?__blob=publicationFile&v=2; Elena Argentesi, Paolo Buccirossi, Emilio Calvano, Tomaso Duso, Alessia Marrazzo, & Salvatore Nava, Merger Policy in Digital Markets: An Ex Post Assessment, 17 Journal of Competition Law & Economics 95 (2021); Marc Bourreau & Alexandre de Streel, Digital Conglomerates and EU Competition Policy (2019), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3350512.

[41] Bundeskartellamt, 11 Feb. 2022, Case B6-21/22, https://www.bundeskartellamt.de/SharedDocs/Entscheidung/EN/Fallberichte/Fusionskontrolle/2022/B6-21-22.html;jsessionid=C0837BD430A8C9C8E04D133B0441EB95.1_cid362?nn=4136442.

[42] UK Competition and Markets Authority, Microsoft / Activision Blizzard Merger Inquiry (2023), https://www.gov.uk/cma-cases/microsoft-slash-activision-blizzard-merger-inquiry.

[43] See European Commission, Commission Prohibits Proposed Acquisition of eTraveli by Booking (2023), https://ec.europa.eu/commission/presscorner/detail/en/ip_23_4573 (finding that a flight product is a crucial growth avenue in Booking’s ecosystem, which revolves around its hotel online-travel-agency (OTA) business, as it would generate significant additional traffic to the platform, thus allowing Booking to benefit from existing customer inertia and making it more difficult for competitors to contest Booking’s position in the hotel OTA market).

[44] Thomas Eisenmann, Geoffrey Parker, & Marshall Van Alstyne, Platform Envelopment, 32 Strategic Management Journal 1270 (2011).

[45] See, e.g., Colangelo, supra note 1, and Pablo Iba?n?ez Colomo, Self-Preferencing: Yet Another Epithet in Need of Limiting Principles, 43 World Competition 417 (2020) (investigating whether and to what extent self-preferencing could be considered a new standalone offense in EU competition law); see also European Commission, Digital Markets Act – Impact Assessment Support Study (2020), 294, https://op.europa.eu/en/publication-detail/-/publication/0a9a636a-3e83-11eb-b27b-01aa75ed71a1/language-en (raising doubts about the novelty of this new theory of harm, which seems similar to the well-established leveraging theories of harm of tying and bundling, and margin squeeze).

[46] European Commission, supra note 45, 16.

[47] European Commission, 27 Jun. 2017, Case AT.39740, Google Search (Shopping).

[48] See General Court, 10 Nov. 2021, Case T-612/17, Google LLC and Alphabet Inc. v. European Commission, ECLI:EU:T:2021:763, para. 155 (stating that the general principle of equal treatment obligates vertically integrated platforms to refrain from favoring their own services as opposed to rival ones; nonetheless, the ruling framed self-preferencing as discriminatory abuse).

[49] In the meantime, however, see Opinion of the Advocate General Kokott, 11 Jan. 2024, Case C-48/22 P, Google v. European Commission, ECLI:EU:C:2024:14, paras. 90 and 95 (arguing that the self-preferencing of which Google is accused constitutes an independent form of abuse, albeit one that exhibits some proximity to cases involving margin squeezing).

[50] European Commission, Commission Sends Amazon Statement of Objections over Proposed Acquisition of iRobot (2023), https://ec.europa.eu/commission/presscorner/detail/en/IP_23_5990.

[51] The same concerns and approach have been shared by the CMA, although it reached a different conclusion, finding that the new merged entity would not have incentive to self-preference its own branded RVCs: see UK Competition and Markets Authority, Amazon / iRobot Merger Inquiry – Clearance Decision (2023), paras. 160, 188, and 231, https://www.gov.uk/cma-cases/amazon-slash-irobot-merger-inquiry.

[52] See European Commission, supra note 45, 304.

[53] Id., 313-314 (envisaging, among potential remedies, the imposition of a duty to make all data used by the platform for strategic decisions available to third parties); see also Désirée Klinger, Jonathan Bokemeyer, Benjamin Della Rocca, & Rafael Bezerra Nunes, Amazon’s Theory of Harm, Yale University Thurman Arnold Project (2020), 19, available at https://som.yale.edu/sites/default/files/2022-01/DTH-Amazon.pdf.

[54] Colangelo, supra note 1; see also Oscar Borgogno & Giuseppe Colangelo, Platform and Device Neutrality Regime: The New Competition Rulebook for App Stores?, 67 Antitrust Bulletin 451 (2022).

[55] See Court of Justice of the European Union (CJEU), 12 May 2022, Case C-377/20, Servizio Elettrico Nazionale SpA v. Autorità Garante della Concorrenza e del Mercato, ECLI:EU:C:2022:379; 19 Apr. 2018, Case C-525/16, MEO v. Autoridade da Concorrência, ECLI:EU:C:2018:270; 6 Sep. 2017, Case C-413/14 P, Intel v. Commission, ECLI:EU:C:2017:632; 6 Oct. 2015, Case C-23/14, Post Danmark A/S v. Konkurrencerådet (Post Danmark II), ECLI:EU:C:2015:651; 27 Mar. 2012, Case C-209/10, Post Danmark A/S v Konkurrencera?det (Post Danmark I), ECLI: EU:C:2012:172; for a recent overview of the EU case law, see also Pablo Iba?n?ez Colomo, The (Second) Modernisation of Article 102 TFEU: Reconciling Effective Enforcement, Legal Certainty and Meaningful Judicial Review, SSRN (2023), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4598161.

[56] CJEU, Intel, supra note 55, paras. 133-134.

[57] CJEU, Servizio Elettrico Nazionale, supra note 55, para. 73.

[58] Opinion of Advocate General Rantos, 9 Dec. 2021, Case C?377/20, Servizio Elettrico Nazionale SpA v. Autorità Garante della Concorrenza e del Mercato, ECLI:EU:C:2021:998, para. 45.

[59] CJEU, Servizio Elettrico Nazionale, supra note 55, para. 77.

[60] Id., paras. 77, 80, and 83.

[61] CJEU, 26 Nov.1998, Case C-7/97, Oscar Bronner GmbH & Co. KG v. Mediaprint Zeitungs- und Zeitschriftenverlag GmbH & Co. KG, Mediaprint Zeitungsvertriebsgesellschaft mbH & Co. KG and Mediaprint Anzeigengesellschaft mbH & Co. KG, ECLI:EU:C:1998:569.

[62] CJEU, Servizio Elettrico Nazionale, supra note 55, para. 85.

[63] European Commission, supra note 11; UK Competition and Markets Authority, supra note 17, paras. 2.6, 4.3, and 4.7.

[64] See, e.g., European Commission, Case COMP D3/34493, DSD, para. 112 (2001) OJ L166/1; affirmed in GC, 24 May 2007, Case T-151/01, DerGru?nePunkt – Duales System DeutschlandGmbH v. European Commission, ECLI:EU:T:2007:154 and CJEU, 16 Jul. 2009, Case C-385/07 P, ECLI:EU:C:2009:456; European Commission, Case IV/31.043, Tetra Pak II, paras. 105–08, (1992) OJ L72/1; European Commission, Case IV/29.971, GEMA III, (1982) OJ L94/12; CJUE, 27 Mar. 1974, Case 127/73, Belgische Radio en Televisie e socie?te? belge des auteurs, compositeurs et e?diteurs v. SV SABAM and NV Fonior, ECLI:EU:C:1974:25, para. 15; European Commission, Case IV/26.760, GEMA II, (1972) OJ L166/22; European Commission, Case IV/26.760, GEMA I, (1971) OJ L134/15.

[65] See, e.g., Richard A. Posner, Intellectual Property: The Law and Economics Approach, 19 The Journal of Economic Perspectives 57 (2005).

[66] See, e.g., Richard Gilbert & Carl Shapiro, Optimal Patent Length and Breadth, 21 The RAND Journal of Economics 106 (1990); Pankaj Tandon, Optimal Patents with Compulsory Licensing, 90 Journal of Political Economy 470 (1982); Frederic M. Scherer, Nordhaus’ Theory of Optimal Patent Life: A Geometric Reinterpretation, 62 American Economic Review 422 (1972); William D. Nordhaus, Invention, Growth, and Welfare: A Theoretical Treatment of Technological Change, Cambridge, MIT Press (1969).

[67] See, e.g., Hal R. Varian, Copying and Copyright, 19 The Journal of Economic Perspectives 121 (2005); William R. Johnson, The Economics of Copying, 93 Journal of Political Economy 158 (1985); Stephen Breyer, The Uneasy Case for Copyright: A Study of Copyright in Books, Photocopies, and Computer Programs, 84 Harvard Law Review 281 (1970).

[68] Sai Krishna Kamepalli, Raghuram Rajan, & Luigi Zingales, Kill Zone, NBER Working Paper No. 27146 (2022), http://www.nber.org/papers/w27146; Massimo Motta & Sandro Shelegia, The “Kill Zone”: Copying, Acquisition and Start-Ups’ Direction of Innovation, Barcelona GSE Working Paper Series Working Paper No. 1253 (2021), https://bse.eu/research/working-papers/kill-zone-copying-acquisition-and-start-ups-direction-innovation; U.S. House of Representatives, Subcommittee on Antitrust, Commercial, and Administrative Law, supra note 8, 164; Stigler Committee for the Study of Digital Platforms, Market Structure and Antitrust Subcommittee (2019) 54, https://research.chicagobooth.edu/stigler/events/single-events/antitrust-competition-conference/digital-platforms-committee; contra, see Geoffrey A. Manne, Samuel Bowman, & Dirk Auer, Technology Mergers and the Market for Corporate Control, 86 Missouri Law Review 1047 (2022).

[69] See also Howard A. Shelanski, Information, Innovation, and Competition Policy for the Internet, 161 University of Pennsylvania Law Review 1663 (2013), 1999 (describing as “forced free riding” the situation occurring when a platform appropriates innovation by other firms that depend on the platform for access to consumers).

[70] See Feng Zhu & Qihong Liu, Competing with Complementors: An Empirical Look at Amazon.com, 39 Strategic Management Journal 2618 (2018).

[71] Andrei Hagiu, Tat-How Teh, and Julian Wright, Should Platforms Be Allowed to Sell on Their Own Marketplaces?, 53 RAND Journal of Economics 297 (2022), (the model assumes that there is a platform that can function as a seller and/or a marketplace, a fringe of small third-party sellers that all sell an identical product, and an innovative seller that has a better product in the same category as the fringe sellers and can invest more in making its product even better; further, the model allows the different channels (on-platform or direct) and the different sellers to offer different values to consumers; therefore, third-party sellers (including the innovative seller) can choose whether to participate on the platform’s marketplace, and whenever they do, can price discriminate between consumers that come to it through the marketplace and consumers that come to it through the direct channel).

[72] See Germa?n Gutie?rrez, The Welfare Consequences of Regulating Amazon (2022), available at http://germangutierrezg.com/Gutierrez2021_AMZ_welfare.pdf (building an equilibrium model where consumers choose products on the Amazon platform, while third-party sellers and Amazon endogenously set prices of products and platform fees).

[73] See Federico Etro, Product Selection in Online Marketplaces, 30 Journal of Economics & Management Strategy 614 (2021), (relying on a model where a marketplace such as Amazon provides a variety of products and can decide, for each product, whether to monetize sales by third-party sellers through a commission or become a seller on its platform, either by commercializing a private label version or by purchasing from a vendor and resell as a first party retailer; as acknowledged by the author, a limitation of the model is that it assumes that the marketplace can set the profit?maximizing commission on each product; if this is not the case, third-party sales would be imperfectly monetized, which would increase the relative profitability of entry).

[74] Patrick Andreoli-Versbach & Joshua Gans, Interplay Between Amazon Store and Logistics, SSRN (2023) https://ssrn.com/abstract=4568024.

[75] Simon Anderson & O?zlem Bedre-Defolie, Online Trade Platforms: Hosting, Selling, or Both?, 84 International Journal of Industrial Organization 102861 (2022).

[76] Chiara Farronato, Andrey Fradkin, & Alexander MacKay, Self-Preferencing at Amazon: Evidence From Search Rankings, NBER Working Paper No. 30894 (2023), http://www.nber.org/papers/w30894.

[77] See Erik Madsen & Nikhil Vellodi, Insider Imitation, SSRN (2023) https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3832712 (introducing a two-stage model where the platform publicly commits to an imitation policy and the entrepreneur observes this policy and chooses whether to innovate: if she chooses not to, the game ends and both players earn profits normalized to zero; otherwise, the entrepreneur pays a fixed innovation cost to develop the product, which she then sells on a marketplace owned by the platform).

[78] Federico Etro, The Economics of Amazon, SSRN (2022), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4307213.

[79] Jay Pil Choi, Kyungmin Kim, & Arijit Mukherjee, “Sherlocking” and Information Design by Hybrid Platforms, SSRN (2023), https://ssrn.com/abstract=4332558 (the model assumes that the platform chooses its referral fee at the beginning of the game and that the cost of entry is the same for both the seller and the platform).

[80] Radostina Shopova, Private Labels in Marketplaces, 89 International Journal of Industrial Organization 102949 (2023), (the model assumes that the market structure is given exogenously and that the quality of the seller’s product is also exogenous; therefore, the paper does not investigate how entry by a platform affects the innovation incentives of third-party sellers).

[81] Jean-Pierre Dube?, Amazon Private Brands: Self-Preferencing vs Traditional Retailing, SSRN (2022) https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4205988.

[82] Gregory S. Crawford, Matteo Courthoud, Regina Seibel, & Simon Zuzek, Amazon Entry on Amazon Marketplace, CEPR Discussion Paper No. 17531 (2022), https://cepr.org/publications/dp17531.

[83] Motta & Shelegia, supra note 68.

[84] Jingcun Cao, Avery Haviv, & Nan Li, The Spillover Effects of Copycat Apps and App Platform Governance, SSRN (2023), https://ssrn.com/abstract=4250292.

[85] Massimo Motta, Self-Preferencing and Foreclosure in Digital Markets: Theories of Harm for Abuse Cases, 90 International Journal of Industrial Organization 102974 (2023).

[86] Id.

[87] Id.

[88] See, e.g., Crawford, Courthoud, Seibel, & Zuzek, supra note 82; Etro, supra note 78; Shopova, supra note 80.

[89] Motta, supra note 85.

[90] Servizio Elettrico Nazionale, supra note 55, paras. 53-54; Post Danmark II, supra note 55, para. 65.

[91] Etro, supra note 78; see also Herbert Hovenkamp, The Looming Crisis in Antitrust Economics, 101 Boston University Law Review 489 (2021), 543, (arguing that: “Amazon’s practice of selling both its own products and those of rivals in close juxtaposition almost certainly benefits consumers by permitting close price comparisons. When Amazon introduces a product such as AmazonBasics AAA batteries in competition with Duracell, prices will go down. There is no evidence to suggest that the practice is so prone to abuse or so likely to harm consumers in other ways that it should be categorically condemned. Rather, it is an act of partial vertical integration similar to other practices that the antitrust laws have confronted and allowed in the past.”)

[92] On the more complex economic rationale of intellectual property, see, e.g., William M. Landes & Richard A. Posner, The Economic Structure of Intellectual Property Law, Cambridge, Harvard University Press (2003).

[93] See, e.g., Italian Competition Authority, 18 Jul. 2023 No. 30737, Case A538 – Sistemi di sigillatura multidiametro per cavi e tubi, (2023) Bulletin No. 31.

[94] See CJEU, 6 Apr. 1995, Joined Cases C-241/91 P and 242/91 P, RTE and ITP v. Commission, ECLI:EU:C:1995:98; 29 Apr. 2004, Case C-418/01, IMS Health GmbH & Co. OHG v. NDC Health GmbH & Co. GH, ECLI:EU:C:2004:257; General Court, 17 Sep. 2007, Case T-201/04, Microsoft v. Commission, ECLI:EU:T:2007:289; CJEU, 16 Jul. 2015, Case C-170/13, Huawei Technologies Co. Ltd v. ZTE Corp., ECLI:EU:C:2015:477.

[95] See, e.g., Dana Mattioli, How Amazon Wins: By Steamrolling Rivals and Partners, Wall Street Journal (2022), https://www.wsj.com/articles/amazon-competition-shopify-wayfair-allbirds-antitrust-11608235127; Aditya Kalra & Steve Stecklow, Amazon Copied Products and Rigged Search Results to Promote Its Own Brands, Documents Show, Reuters (2021), https://www.reuters.com/investigates/special-report/amazon-india-rigging.

[96] Williams-Sonoma, Inc. v. Amazon.Com, Inc., Case No. 18-cv-07548 (N.D. Cal., 2018). The suit was eventually dismissed, as the parties entered into a settlement agreement: Williams-Sonoma, Inc. v. Amazon.Com, Inc., Case No. 18-cv-07548-AGT (N.D. Cal., 2020).

[97] Amazon Best Sellers, https://www.amazon.com/Best-Sellers/zgbs.

[98] Hovenkamp, supra note 91, 2015-2016.

[99] Nicolas Petit, Big Tech and the Digital Economy, Oxford, Oxford University Press (2020), 224-225.

[100] For a recent analysis, see Zijun (June) Shi, Xiao Liu, Dokyun Lee, & Kannan Srinivasan, How Do Fast-Fashion Copycats Affect the Popularity of Premium Brands? Evidence from Social Media, 60 Journal of Marketing Research 1027 (2023).

[101] Lina M. Khan, Amazon’s Antitrust Paradox, 126 Yale Law Journal 710 (2017), 782.

[102] See Massimo Motta &Martin Peitz, Intervention Triggers and Underlying Theories of Harm, in Market Investigations. A New Competition Tool for Europe? (M. Motta, M. Peitz, & H. Schweitzer, eds.), Cambridge, Cambridge University Press (2022), 16, 59 (arguing that, while it is unclear to what extent products or ideas are worth protecting and/or can be protected from sherlocking and whether such cloning is really harmful to consumers, this is clearly an area where an antitrust investigation for abuse of dominant position would not help).

[103] Khan, supra note 101, 780 and 783 (arguing that Amazon’s conflicts of interest tarnish the neutrality of the competitive process and that the competitive implications are clear, as Amazon is exploiting the fact that some of its customers are also its rivals).

[104] Servizio Elettrico Nazionale, supra note 55, para. 85.

[105] Post Danmark I, supra note 55, para. 22.

[106] Iba?n?ez Colomo, supra note 55, 21-22.

[107] Id.

[108] See, e.g., DMA, supra note 4, Recital 5 (complaining that the scope of antitrust provisions is “limited to certain instances of market power, for example dominance on specific markets and of anti-competitive behaviour, and enforcement occurs ex post and requires an extensive investigation of often very complex facts on a case by case basis.”).

[109] U.S. Federal Trade Commission, et al. v. Amazon.com, Inc., supra note 23.

[110] Khan, supra note 101.

[111] Khan, supra note 22, 1003, referring to Amazon, Google, and Meta.

Continue reading
Antitrust & Consumer Protection

Blackout Rebates: Tipping the Scales at the FCC

TOTM Cable and satellite programming blackouts often generate significant headlines. While the share of the population affected by blackouts may be small—bordering on minuscule—most consumers don’t . . .

Cable and satellite programming blackouts often generate significant headlines. While the share of the population affected by blackouts may be small—bordering on minuscule—most consumers don’t like the idea of programming blackouts and balk at the idea of paying for TV programming they can’t access.

Read the full piece here.

Continue reading
Telecommunications & Regulated Utilities

Questions Arise on SB 1596: The Right to Repair Bill

Popular Media The Oregon Senate earlier this month approved SB 1596, the so-called “right to repair” bill. This legislation now awaits consideration in the Oregon House, with . . .

The Oregon Senate earlier this month approved SB 1596, the so-called “right to repair” bill. This legislation now awaits consideration in the Oregon House, with a hearing of the House Committee on Business and Labor scheduled for Wednesday.

While motivated by good intentions, this legislation risks unintended consequences that could ultimately harm consumers. Lawmakers should proceed cautiously.

Read the full piece here.

Continue reading
Intellectual Property & Licensing

Congress Should Protect the Rights of American Creators with Site-Blocking Legislation

Scholarship Summary Large-scale piracy websites violate the copyrights of American creators and threaten the continued growth of the creative industries. The 1998 Digital Millennium Copyright Act . . .

Summary

Large-scale piracy websites violate the copyrights of American creators and threaten the continued growth of the creative industries. The 1998 Digital Millennium Copyright Act is ineffective in stopping this scourge, as its protections are limited to obsolete technologies and it does not apply to piracy websites based in foreign countries. Congress should protect the rights of American creators on the 21st-century internet by enacting site-blocking legislation. Many U.S. allies already have site-blocking laws. A decade of studies and data from the operation of these site-blocking laws have proven these laws work without chilling speech or “breaking the internet.” Site-blocking laws are a proven, effective mechanism in protecting copyrights and promoting legitimate online commercial services.
Continue reading
Intellectual Property & Licensing

Kristian Stout on Artificial Intelligence and Copyright

Presentations & Interviews 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 . . .

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.

Continue reading
Intellectual Property & Licensing

ICLE Comments on Artificial Intelligence and Copyright

Regulatory Comments 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 . . .

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 (9th Cir. 2002), the 9th 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 9th 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 9th 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 4th 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 4th 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 5th 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 (9th 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 20th 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 (9th 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.

Continue reading
Intellectual Property & Licensing