TLDR: Federal judges have ruled that Meta and Anthropic’s use of copyrighted books to train their AI models constitutes fair use, even when some materials were sourced from unauthorized sites. The decisions, handed down by Judge Vince Chhabria for Meta and Judge William Alsup for Anthropic, emphasize the transformative nature of AI training and the plaintiffs’ failure to demonstrate direct market harm, though future cases may hinge on a new ‘market dilution’ theory.
In a significant development for the burgeoning artificial intelligence industry, two separate U.S. federal courts have delivered rulings affirming that the use of copyrighted literary works to train generative AI models falls under the doctrine of fair use. These decisions, issued on June 26, 2025, provide crucial early precedents in the ongoing legal battles between content creators and AI developers.
In the case of Kadrey v. Meta Platforms, Judge Vince Chhabria ruled in favor of Meta, stating that the company did not violate copyright laws by using books from 13 authors, including prominent figures like Sarah Silverman and Ta-Nehisi Coates, to train its large language models (LLMs). Judge Chhabria’s decision primarily hinged on the plaintiffs’ inability to present sufficient evidence that Meta’s use of their work caused direct financial harm or substantially diminished the market for their original creations. ‘The Court has no choice but to grant summary judgment to Meta,’ Judge Chhabria wrote, emphasizing that ‘The key question in virtually any case where a defendant has copied someone’s original work without permission is whether allowing people to engage in that sort of conduct would substantially diminish the market for the original.’
While ruling for Meta, Judge Chhabria introduced a novel concept of ‘market dilution,’ suggesting that using copyrighted books to train an LLM ‘might harm the market for those works because it enables the rapid generation of countless works that compete with the originals, even if those works aren’t themselves infringing.’ This theory, a departure from traditional ‘market substitution,’ could become a critical argument for future plaintiffs seeking to prove harm.
Analyzing the four factors of fair use, the court found that Meta’s use was highly transformative (Factor 1), as LLMs serve different functions than the original books. While the nature of the copyrighted works (Factor 2) favored the authors, the amount copied (Factor 3) — even entire books — was deemed ‘reasonably necessary’ for effective AI training, thus favoring Meta. The crucial fourth factor, market effect, ultimately favored Meta due to the plaintiffs’ insufficient evidence of harm.
This ruling follows closely on the heels of a similar decision in Bartz v. Anthropic, where Judge William Alsup also found Anthropic’s AI training practices to be fair use. Judge Alsup’s reasoning placed a stronger emphasis on the ‘spectacularly transformative’ nature of AI. However, a key distinction in the Anthropic case is that Judge Alsup ordered a separate trial to address whether Anthropic infringed copyrights by storing pirated books, indicating that while training might be permissible, the acquisition and retention methods of data could still lead to liability.
Meta spokesperson Thomas Richards welcomed the decision, stating, ‘Open-source AI models are powering transformative innovations, productivity, and creativity for individuals and companies, and fair use of copyright material is a vital legal framework for building this transformative technology.’
Conversely, the plaintiffs’ legal representatives expressed disappointment, arguing that despite the court acknowledging that AI companies ‘feed copyright-protected works into their models without getting permission’ generally violate the law, the ruling still favored Meta. They suggested the authors ‘made the wrong arguments and failed to develop a record in support of the right one,’ particularly regarding market harm.
Also Read:
- Authors File Lawsuit Against Microsoft Over Alleged Use of Pirated Books for AI Training
- UK Courts Grapple with AI Misuse: Lawyers Referred to Regulators Over Fabricated Case Citations
These district court decisions, which are appealable, represent some of the first judicial interpretations of fair use in the context of generative AI. They underscore the importance for future plaintiffs to build a robust evidentiary record demonstrating direct market impact or dilution to successfully challenge AI companies’ use of copyrighted materials. The rulings are expected to significantly influence the legal strategies in numerous other copyright lawsuits currently pending against AI developers worldwide.
