Introduction
Recent rulings by Judges William Alsup and Vince Chhabria have addressed the complex issue of fair use in copyright infringement cases involving book authors and technology companies developing generative AI large language models (LLMs). These decisions [1] [2] [3] [5] [7], which involve companies such as Anthropic and Meta Platforms Inc, have significant implications for the use of copyrighted materials in AI training.
Description
Judges William Alsup and Vince Chhabria have recently issued significant rulings in copyright infringement lawsuits involving book authors and technology companies developing generative AI large language models (LLMs), specifically Anthropic and Meta Platforms Inc. Both judges determined that the use of copyrighted books for training LLMs constituted fair use, thereby not infringing copyright [2]. Judge Alsup ruled in favor of Anthropic [5], affirming that its use of copyrighted materials was transformative [5], creating new content rather than replacing the originals [6]. He expressed concerns regarding Anthropic’s storage of over 7 million pirated books in a central library for training its Claude models, which could lead to copyright infringement [1]. A separate trial is anticipated in December 2025 to address potential damages owed by Anthropic.
Meanwhile, Judge Chhabria found that Meta’s use of copyrighted books for training its LLaMA models also qualified as fair use. He emphasized that the plaintiffs [2], including notable authors like Sarah Silverman, Ta-Nehisi Coates [3], and Junot Díaz [3], failed to demonstrate any significant market harm to their works, suggesting that their case was not strong enough to challenge Meta’s practices. Chhabria characterized the use of the authors’ content as “highly transformative,” serving a different purpose than the original texts [1]. He noted that the quality of an LLM is contingent on the breadth and quality of the training materials [2], justifying the necessity of copying entire books [2]. However, he cautioned that evidence of market harm could influence future rulings [7], particularly as the judges diverged on a new market dilution theory that could challenge the tech companies’ fair use defenses in subsequent litigation [2].
Both judges dismissed the authors’ claims that the AI companies had replaced the market for licensing their works [2], despite evidence that both companies had explored licensing options [2]. Alsup argued that licensing for the specific purpose of training LLMs is not a right protected by the Copyright Act [2], while Chhabria highlighted the challenges Meta faced in securing licenses [2], including issues with rights ownership and publisher responsiveness [2]. Additionally, both companies are facing ongoing legal challenges regarding allegations of illegally obtaining copyrighted materials from pirated sources [6].
Judge Alsup emphasized that copyright law only protects against substantially similar copying and not against all forms of competition [4], indicating that competitive displacement without substantial similarity falls outside copyright’s scope [4]. In contrast, Judge Chhabria’s assumption that copyright addresses all creative sector issues overlooks the Act’s limitation to imitative copying [4]. Concerns about human creative displacement raised by Chhabria fall outside copyright law’s purview and may be better addressed through alternative policy measures [4], such as a tax on Generative AI development to support human authors [4].
Looking ahead, Judge Eumi Lee is expected to rule on fair use in the Concord Music v [7]. Anthropic case [1] [2] [6] [7] [8], with a decision anticipated by the end of 2025 [7]. The implications of the recent fair use decisions by Judges Alsup and Chhabria are still being assessed [7], and Judge Lee’s ruling will further clarify whether the use of copyrighted works for AI training constitutes fair use [7]. These rulings may have significant repercussions for ongoing lawsuits against various AI companies, particularly in sectors like news publishing [5], where the competitive threat from AI tools is increasingly evident. The outcomes of these cases highlight the complexity of fair use determinations, which are highly fact-specific and may evolve as new challenges arise, particularly concerning the use of pirated material and the potential impact on the market.
Conclusion
The rulings by Judges Alsup and Chhabria underscore the evolving nature of copyright law in the context of AI development. Their decisions highlight the nuanced considerations involved in determining fair use, particularly as it pertains to transformative use and market impact. As the legal landscape continues to develop, these cases may set important precedents for future litigation involving AI and copyrighted materials, influencing both legal strategies and policy discussions in the creative and technological sectors.
References
[1] https://worldlawyersforum.org/news/meta-anthropic-ai-copyright-win/
[2] https://news.bloomberglaw.com/us-law-week/big-tech-wins-in-copyright-cases-come-with-strings-attached
[3] https://usaherald.com/judge-rules-metas-ai-use-of-books-is-fair-use/
[4] https://spicyip.com/2025/07/part-ii-obiter-ratio-the-good-and-bad-in-the-first-two-american-decisions-on-generative-ai-and-copyright.html
[5] https://observer.com/2025/06/meta-anthropic-fair-use-wins-ai-copyright-cases/
[6] https://www.technologyreview.com/2025/07/01/1119486/ai-copyright-meta-anthropic/
[7] https://chatgptiseatingtheworld.com/2025/07/07/which-judge-will-decide-fair-use-next-in-ai-copyright-litigation-judge-eumi-lee-most-likely/
[8] https://www.techtarget.com/searchenterpriseai/news/366626775/Fair-use-rulings-favor-Meta-and-Anthropic-but-are-limited
