Anthropic has agreed to pay $1.5 billion to settle a class-action copyright lawsuit brought by authors who accused the AI company of using pirated copies of their books to train its Claude language model. The proposed settlement, which requires court approval, would distribute roughly $3,000 per work across an estimated 500,000 books; the motion seeking approval calls it the “largest publicly reported copyright recovery in history.”

The case, Bartz v. Anthropic, was set to go to trial in December. In July, U.S. District Judge William Alsup ruled that training AI on lawfully acquired books qualifies as fair use, but he allowed the case to proceed on the narrower claim that Anthropic had downloaded books from shadow libraries such as Library Genesis. That distinction exposed the company to potentially ruinous statutory damages — some estimates ran as high as $1 trillion if willful infringement were found.

Anthropic, which this week announced a $13 billion fundraising round at a $183 billion valuation, said it felt ‘inordinate pressure’ to settle. The company will also destroy the pirated files it was accused of amassing. Legal experts said the deal does not set a binding precedent, but it provides a reference point for similar cases pending against OpenAI, Meta, and others — and, some argued, a nudge toward a licensed, market-based model for training data.