Consent and Compensation: Resolving Generative AI’s Copyright Crisis

*Frank Pasquale is a Professor of Law at Cornell Law School and Cornell Tech. Haochen Sun is a Professor of Law at the University of Hong Kong Faculty of Law. We thank Shyam Balganesh, Anupam Chander, William Fisher, James Grimmelmann, Jacob Noti-Victor, Ben Sobel, Scott Veitch, and Christopher Yoo for valuable comments and conversations. We are grateful to participants in the Hong Kong University conference “Reframing Intellectual Property Law in the Age of Artificial Intelligence” for their comments in response to a presentation of this project. We also thank Michelle Brodsky, Alex Cho, Jae Shin, and Upasana Singh for excellent research assistance.Show More

Generative artificial intelligence (AI) has the potential to augment and democratize creativity. However, it is undermining the knowledge ecosystem that now sustains it. Generative AI may unfairly compete with authors, journalists, and other creative workers, displacing them in the market. Most AI firms are not compensating creative workers for composing the songs, drawing the images, and writing both the fiction and nonfiction books that their models need in order to function. AI thus threatens not only to undermine the livelihoods of authors, artists, and other creatives, but also to destabilize the very knowledge ecosystem it relies on.

Alarmed by these developments, many copyright owners have objected to the use of their works by AI providers. In order to recognize and empower their demands to stop nonconsensual use of their works, we propose a streamlined opt-out mechanism that would require AI providers to remove objectors’ works from their databases once copyright infringement has been documented. Those who do not object still deserve compensation for the use of their work by AI providers. We thus also propose a levy on AI providers, to be distributed to the copyright owners whose work they use without a license. This scheme is designed to ensure that creatives receive a fair share of the economic bounty arising out of their contributions to AI. Together, these mechanisms of consent and compensation would result in a new grand bargain between copyright owners and AI firms, helping to ensure the long-term viability of both AI and the human thought and expression it depends on.

Introduction

From the printing press to the Internet, technological advance has profoundly changed the way authors create, disseminate, and monetize their works.¹ 1.See generally Adrian Johns, Piracy: The Intellectual Property Wars from Gutenberg to Gates(2009) (discussing the history of copyright piracy).Show More Widespread access to the Internet has caused book, music, and film creators great economic setbacks via piracy, but has also created new opportunities, particularly for “long tail” creators shunned by dominant recording companies and broadcasters.² 2.See Chris Anderson, The Long Tail, Wired (Oct. 1, 2004, 12:00 PM), https://www.wired.‌com/2004/10/tail [https://perma.cc/P9QQ-MPTG].Show More Despite the upheaval, human authors have remained indispensable in the creation of works, as pirates do not create original content.

The rise of generative artificial intelligence (AI), however, represents an inflection point.³ 3.Generative AI’s power to create exact replicas of existing works, and to imitate many characteristic elements of existing works, has provoked a wave of lawsuits over the past two years. However, copyright controversies over the training of AI antedate the rise of generative AI. To mark the relevance of that past work, and the continuity of the problems likely to be raised by AI when the next generation of AI arises, we refer to “AI” throughout the Essay, rather than the more cumbersome “generative AI” or “GenAI.”Show More AI can plagiarize at a far faster rate than human copyists.⁴ 4.Kate Knibbs, Scammy AI-Generated Book Rewrites Are Flooding Amazon, Wired (Jan. 10, 2024, 7:00 AM), https://www.wired.com/story/scammy-ai-generated-books-flooding-am‌azon/ [https://perma.cc/4R7G-LXFU].Show More These capacities are menacing both fiction and nonfiction book authors, as well as journalists.⁵ 5.Our focus in this Essay is on corporations developing, marketing, and selling AI services. The legislative approaches developed in this Essay may, in a calibrated fashion, adjust duties of AI providers to reflect their size, for-profit or nonprofit status, and other factors.Show More AI can also create new works that closely resemble the style and content of existing ones. When prompted skillfully, large language models (LLMs) aid in the rapid creation of a high volume of content. The bottom line is an “existential crisis” for many creatives, threatening to drive the marginal value of their labor below subsistence levels as cheap AI content displaces human works.⁶ 6.See Michael Cavna, Artists Are Alarmed by AI—and They’re Fighting Back, Wash. Post (Feb. 14, 2023, 6:00 AM), https://www.washingtonpost.com/comics/2023/02/14/ai-in-illustra‌tion/ [https://perma.cc/4RFW-5FX3] (describing “an existential threat to the livelihood of artists”). Throughout this Essay, we will refer to artists, writers, journalists, and other creators of expressive works as “creatives” or “copyright owners.” We realize these terms may be too capacious: some expressive work only takes a minimal amount of creativity, and many creatives have transferred their copyrights to others in exchange for compensation. Nevertheless, copyright is premised on some minimal level of creativity, and the future compensation of creatives who plan to alienate their copyrights is at least in part premised on the value of those copyrights to those seeking them. Thus the terms capture enough of social and economic reality to be useful here.Show More

Given the enthusiasm for AI evident among so many owners of dominant content distribution platforms, such a displacement may already be underway.⁷ 7.Edward Zitron, Are We Watching the Internet Die?, Where’s Your Ed At? (Mar. 11, 2024), https://www.wheresyoured.at/are-we-watching-the-internet-die/ [https://perma.cc/PZC5-H9‌FF] (recognizing that because “platforms were built to reward scale and volume far more often than quality,” creatives who use AI enjoy important advantages over those who do not).Show More To create and improve their AI models, large technology firms have undermined authors’ proprietary control over their works by using these works as training data, without consent and often through opaque processes.⁸ 8.See infra Section I.B.Show More At the same time, AI systems like ChatGPT and MidJourney can rapidly generate a wide variety of content, potentially outperforming humans in the marketplace of ideas—particularly when so many of this marketplace’s main organizers such as Alphabet (Google’s parent company), X (formerly Twitter), and Meta (formerly Facebook) are themselves developing AI.⁹ 9.See Thomas H. Davenport & Nitin Mittal, How Generative AI Is Changing Creative Work, Harv. Bus. Rev. (Nov. 14, 2022), https://hbr.org/2022/11/how-generative-ai-is-changing-crea‌tive-work [https://perma.cc/SK98-ZE5T].Show More

To compound these challenges, leading firms in the AI space are unlikely to offer compensation for the vital contributions of copyrighted works to their systems. In 2023, this state of affairs helped lead to an unprecedented 148-day strike by Hollywood screenwriters.¹⁰ 10.Ben Schwartz, AI and the Hollywood Writers’ Strike, Nation (May 8, 2023), https://www.‌thenation.com/article/economy/ai-and-the-hollywood-writers-strike [https://perma.cc/8TJR-ZBUC]; Jennifer Maas, The Writers Strike Is Over: WGA Votes to Lift Strike Order After 148 Days, Variety (Sept. 26, 2023, 5:07 PM), https://variety.com/2023/tv/news/writers-strike-over-wga-votes-end-work-stoppage-1235735512/ [https://perma.cc/F5P7-QEWF].Show More Book authors are also alarmed. Over 15,000 writers, including prominent novelists such as Dan Brown, Suzanne Collins, and Margaret Atwood, have endorsed an open letter demanding fair compensation, credit, and author consent for the use of their works in AI systems.¹¹ 11.Open Letter to Generative AI Leaders, Action Network, https://actionnetwork.org/petitio‌ns/authors-guild-open-letter-to-generative-ai-leaders [https://perma.cc/8D5W-WGFL] (last visited Mar. 3, 2024).Show More At least one former executive in an AI firm has resigned his position, considering the unlicensed use of music as training data both ethically and legally untenable.¹² 12.Kate Knibbs, This Tech Exec Quit His Job to Fight Generative AI’s Original Sin, Wired (Jan. 17, 2024, 4:44 PM), https://www.wired.com/story/ai-executive-ed-newton-rex-turns-cru‌sader-stand-up-for-artists [https://perma.cc/97NE-H4Y7].Show More This struggle has resulted in numerous courtroom battles over copyright infringement, too.¹³ 13.Complaint at 2–3, Basbanes v. Microsoft Corp., No. 24-cv-00084 (S.D.N.Y. Jan. 5, 2024); Complaint at 2–4, N.Y. Times Co. v. Microsoft Corp., No. 23-cv-11195 (S.D.N.Y. Dec. 27, 2023); Generative AI-Intellectual Property Cases and Policy Tracker, Mishcon de Reya LLP, https://www.mishcon.com/generative-ai-intellectual-property-cases-and-policy-tr‌acker [https://perma.cc/7RHU-3PG2] (last visited Mar. 3, 2024).Show More AI firms claim that they are protected by the fair use defense,¹⁴ 14.Mark A. Lemley & Bryan Casey, Fair Learning, 99 Tex. L. Rev. 743, 748 (2021) (arguing that “a [machine learning] system’s use of the data often is transformative as that term has come to be understood in copyright law, because even though it doesn’t change the underlying work, it changes the purpose for which the work is used”).Show More but application of the doctrine is notoriously uncertain, particularly with respect to new technologies.¹⁵ 15.Katherine Lee, A. Feder Cooper & James Grimmelmann, Talkin’ ‘Bout AI Generation: Copyright and the Generative-AI Supply Chain, 71 J. Copyright Soc’y (forthcoming 2024) (manuscript at 105), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4523551 [https://pe‌rma.cc/Z3C7-PJWJ] (“[F]air use is famously case-specific, so no ex ante analysis can anticipate all of the relevant issues.”).Show More

This litigation may drag on for years, slowing the development of AI while denying or delaying fair compensation to creatives. The situation strikes many policymakers as deeply unfair and undesirable. As the Communications and Digital Committee of the United Kingdom’s House of Lords has concluded, “[w]e do not believe it is fair for tech firms to use rightsholder data for commercial purposes without permission or compensation, and to gain vast financial rewards in the process.”¹⁶ 16.Commc’ns & Digit. Comm., Large Language Models and Generative AI, 2023-24, HL 54, ¶ 245 (UK).Show More A legislative solution is desirable, and there is a venerable tradition of actual and proposed solutions to the copyright problems created by new technological uses of works.¹⁷ 17.See William W. Fisher III, Promises to Keep: Technology, Law, and the Future of Entertainment 1–22 (2004).Show More

To guide policymakers, this Essay outlines a promising framework for a legislative solution premised on coupling mechanisms of control (via opt-out rights) and compensation (via a levy to be imposed on AI providers by a central authority and then distributed to owners of works used by those AI providers without a license). These mechanisms could first be imposed on the largest AI providers and then expanded as appropriate once standardized. Part I explains the urgency of this proposal by demonstrating that free expropriation of copyrighted works by AI providers not only devalues human creativity, but also threatens to undermine AI itself by eliminating critical incentives for the ongoing creation of works necessary for further technological development. Part II outlines an opt-out mechanism, permitting creatives to forbid nonconsensual use of their works for training AI models after documenting copyright infringement. Part III addresses the proper level of levies necessary to compensate those who do not choose to opt out or license their works to AI providers. Part IV anticipates and responds to objections to our proposal. This Essay concludes by reflecting on the broader policy implications of our proposal.

1.  See generally Adrian Johns, Piracy: The Intellectual Property Wars from Gutenberg to Gates
   

   (2009) (discussing the history of copyright piracy). ↑

2.  See Chris Anderson, The Long Tail, Wired (Oct. 1, 2004, 12:00 PM), https://www.wired.‌com/2004/10/tail [https://perma.cc/P9QQ-MPTG]. ↑
3.  Generative AI’s power to create exact replicas of existing works, and to imitate many characteristic elements of existing works, has provoked a wave of lawsuits over the past two years. However, copyright controversies over the training of AI antedate the rise of generative AI. To mark the relevance of that past work, and the continuity of the problems likely to be raised by AI when the next generation of AI arises, we refer to “AI” throughout the Essay, rather than the more cumbersome “generative AI” or “GenAI.” ↑
4.  Kate Knibbs, Scammy AI-Generated Book Rewrites Are Flooding Amazon, Wired (Jan. 10, 2024, 7:00 AM), https://www.wired.com/story/scammy-ai-generated-books-flooding-am‌azon/ [https://perma.cc/4R7G-LXFU]. ↑
5.  Our focus in this Essay is on corporations developing, marketing, and selling AI services. The legislative approaches developed in this Essay may, in a calibrated fashion, adjust duties of AI providers to reflect their size, for-profit or nonprofit status, and other factors. ↑
6.  See Michael Cavna, Artists Are Alarmed by AI—and They’re Fighting Back, Wash. Post
   

   (

   Feb. 14, 2023, 6:00 AM), https://www.washingtonpost.com/comics/2023/02/14/ai-in-illustra‌tion/ [https://perma.cc/4RFW-5FX3] (describing “an existential threat to the livelihood of artists”). Throughout this Essay, we will refer to artists, writers, journalists, and other creators of expressive works as “creatives” or “copyright owners.” We realize these terms may be too capacious: some expressive work only takes a minimal amount of creativity, and many creatives have transferred their copyrights to others in exchange for compensation. Nevertheless, copyright is premised on some minimal level of creativity, and the future compensation of creatives who plan to alienate their copyrights is at least in part premised on the value of those copyrights to those seeking them. Thus the terms capture enough of social and economic reality to be useful here. ↑

7.  Edward Zitron, Are We Watching the Internet Die?, Where’s Your Ed At? (Mar. 11, 2024), https://www.wheresyoured.at/are-we-watching-the-internet-die/ [https://perma.cc/PZC5-H9‌FF] (recognizing that because “platforms were built to reward scale and volume far more often than quality,” creatives who use AI enjoy important advantages over those who do not). ↑
8.  See infra Section I.B. ↑
9.  See Thomas H. Davenport & Nitin Mittal, How Generative AI Is Changing Creative Work, Harv. Bus. Rev. (Nov. 14, 2022), https://hbr.org/2022/11/how-generative-ai-is-changing-crea‌tive-work [https://perma.cc/SK98-ZE5T]. ↑
10.  Ben Schwartz, AI and the Hollywood Writers’ Strike, Nation (May 8, 2023), https://www.‌thenation.com/article/economy/ai-and-the-hollywood-writers-strike [https://perma.cc/8TJR-ZBUC]; Jennifer Maas, The Writers Strike Is Over: WGA Votes to Lift Strike Order After 148 Days, Variety (Sept. 26, 2023, 5:07 PM), https://variety.com/2023/tv/news/writers-strike-over-wga-votes-end-work-stoppage-1235735512/ [https://perma.cc/F5P7-QEWF]. ↑
11.  Open Letter to Generative AI Leaders, Action Network, https://actionnetwork.org/petitio‌ns/authors-guild-open-letter-to-generative-ai-leaders [https://perma.cc/8D5W-WGFL] (last visited Mar. 3, 2024). ↑
12.  Kate Knibbs, This Tech Exec Quit His Job to Fight Generative AI’s Original Sin, Wired
    

    (

    Jan. 17, 2024, 4:44 PM), https://www.wired.com/story/ai-executive-ed-newton-rex-turns-cru‌sader-stand-up-for-artists [https://perma.cc/97NE-H4Y7]. ↑

13.  Complaint at 2–3, Basbanes v. Microsoft Corp., No. 24-cv-00084 (S.D.N.Y. Jan. 5, 2024); Complaint at 2–4, N.Y. Times Co. v. Microsoft Corp., No. 23-cv-11195 (S.D.N.Y. Dec. 27, 2023); Generative AI-Intellectual Property Cases and Policy Tracker, Mishcon de Reya LLP, https://www.mishcon.com/generative-ai-intellectual-property-cases-and-policy-tr‌acker [https://perma.cc/7RHU-3PG2] (last visited Mar. 3, 2024). ↑
14.  Mark A. Lemley & Bryan Casey, Fair Learning, 99 Tex. L. Rev. 743, 748 (2021) (arguing that “a [machine learning] system’s use of the data often is transformative as that term has come to be understood in copyright law, because even though it doesn’t change the underlying work, it changes the purpose for which the work is used”). ↑
15.  Katherine Lee, A. Feder Cooper & James Grimmelmann, Talkin’ ‘Bout AI Generation: Copyright and the Generative-AI Supply Chain, 71 J. Copyright Soc’y (forthcoming 2024) (manuscript at 105), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4523551 [https://pe‌rma.cc/Z3C7-PJWJ] (“[F]air use is famously case-specific, so no ex ante analysis can anticipate all of the relevant issues.”). ↑
16.  Commc’ns & Digit. Comm., Large Language Models and Generative AI, 2023-24, HL 54, ¶ 245 (UK). ↑

17.  See

    William W. Fisher III, Promises to Keep: Technology, Law, and the Future of Entertainment 1–22 (2004). ↑