U.S. judge rules AI training on copyrighted books lawful in $1.5B settlement, complicating fair-use debate

A 2025 ruling by U.S. Judge William Alsup found that training AI models on copyrighted books is lawful under copyright law, despite a $1.5 billion settlement against Anthropic for obtaining books from illegal online shadow libraries. The judge compared the way large language models ingest trillions of words to a writer’s study of literature, emphasizing that AI training is akin to reading or consuming a work rather than copying it.

Legal experts interviewed by TechCrunch described the ruling as advantageous for AI companies, noting that the $1.5 billion penalty was tied to the method of acquisition—piracy—not the act of training itself. Anthropic has projected annual revenue of about $200 billion by 2028, making the fine a fraction of its projected scale.

The decision leaves unresolved broader questions about fair use in AI training, particularly when AI outputs could compete with original works. Courts have historically considered four factors in fair-use determinations: the purpose and character of the use, the nature of the copyrighted work, the amount used, and the effect on the market. Experts say outcomes hinge on whether the AI’s purpose is transformative or competitive.

In a separate case, Thomson Reuters sued Ross Intelligence for copying its content to build a competing AI-based legal platform. The court ruled that Ross’s use was not transformative because it lacked a further purpose or different character than the original, signaling that courts may reject fair-use claims when AI systems directly compete with copyright holders.

Copyright law has not been updated since 1976, forcing judges to interpret decades-old guidelines in the context of AI. Attorneys and legal scholars told TechCrunch that the lack of clarity is shaping industry behavior, with companies making litigation-risk decisions based on early, inconsistent rulings.

The ruling is one of the first of its kind and is expected to influence ongoing and future litigation involving AI companies and copyright holders. Legal experts cautioned that early rulings are not final and may be overturned or distinguished in later cases, prolonging uncertainty for the industry.