A federal judge has approved a $1.5 billion settlement requiring Anthropic to compensate thousands of authors approximately $3,000 per book after it was found to have used pirated copies of copyrighted works to train its Claude chatbot. The ruling, issued Monday by Judge Araceli Martínez-Olguín, was part of a class-action lawsuit involving more than 482,000 books.

Judge Martínez-Olguín described the settlement as providing “meaningful relief” to the authors and publishers affected by Anthropic’s unauthorized use of their works. According to court records, about 91 percent of the books covered by the settlement have already been claimed by their rights holders, who will now receive payments as part of the agreement.

Justin Nelson, lead counsel for the plaintiffs, characterized the settlement as “the largest known copyright recovery in history” and indicated that efforts are underway to distribute the funds to the class members “as promptly as possible.”

The case stems from allegations that Anthropic, an artificial intelligence research company, trained its Claude chatbot using copyrighted books obtained from pirate websites without permission. While U.S. District Judge William Alsup ruled last summer that training AI models on copyrighted material was not inherently unlawful, he found that Anthropic unlawfully acquired those materials through illicit means.

The settlement agreement marks a significant development in ongoing debates over AI training practices and copyright law, reflecting legal scrutiny of how AI companies source protected content. It remains a notable example of the challenges posed by the rapid expansion of AI technologies and their reliance on vast datasets.