Two federal judges in California ruled within days of each other in June 2025 that using copyrighted books to train generative AI models can be fair use, a significant test case outcome for the AI industry as it faces a wave of copyright lawsuits from authors, publishers and news organizations.
In Bartz v. Anthropic, Judge William Alsup ruled that Anthropic's use of copyrighted books to train its Claude models was fair use, but ordered the company to face a separate trial over claims it pirated millions of books from unauthorized sources rather than paying for them. Days later, Judge Vince Chhabria granted summary judgment to Meta in Kadrey v. Meta Platforms, finding the plaintiffs had not shown that Meta's use of their books to train its Llama models caused the kind of market harm needed to defeat a fair-use defense.
Why it matters
These were among the first substantive fair-use rulings in the wave of AI copyright litigation filed against OpenAI, Meta, Anthropic, Stability AI, Microsoft and others since 2023. Because so many other cases raise similar fair-use questions, the reasoning in Bartz and Kadrey is likely to influence how other judges — and eventually appeals courts — approach the core question of whether training a model on copyrighted text without a license is lawful.
How the rule works
Fair use under U.S. copyright law, codified at 17 U.S.C. Section 107, weighs four factors: the purpose and character of the use, the nature of the copyrighted work, the amount used, and the effect on the market for the original. Both judges found the “transformative” nature of using text to train a model — rather than to reproduce or compete with the original works in the market — weighed heavily in favor of the AI companies. But Judge Alsup drew a sharp distinction between the training use itself, which he found transformative, and the separate question of how Anthropic obtained the books, ruling that downloading pirated copies could constitute copyright infringement independent of any later fair-use training analysis.
Evidence
Judge Chhabria's June 25, 2025 order in Kadrey v. Meta Platforms, Case No. 23-cv-03417-VC, granted Meta's cross-motion for partial summary judgment on the fair-use question while denying the plaintiffs' motion, but Chhabria's opinion itself cautioned that the ruling was narrow and reflected weaknesses in how these particular plaintiffs argued market harm, not a blanket endorsement of AI training practices.
The competing read
AI companies and their supporters have pointed to both rulings as validating the industry's long-standing argument that training on copyrighted material is transformative and does not require licensing every source text. Authors' groups and copyright holders counter that the rulings are narrower than headlines suggest — as legal analysts at Debevoise & Plimpton and TechTarget noted, the decisions turned on case-specific facts and evidentiary gaps, not a categorical rule that AI training is always fair use, and the piracy claims against Anthropic show real liability risk remains over sourcing.
What happens next
Anthropic still faced a separate trial over the piracy allegations following the June 2025 ruling, a proceeding being watched closely because it could determine damages exposure even after the fair-use question was resolved in Anthropic's favor. Numerous other AI copyright cases, including suits against OpenAI and Microsoft brought by The New York Times and other publishers, remain pending, and legal commentators expect the fair-use question to eventually reach a federal appeals court and potentially the Supreme Court.