Federal judge greenlights $1.5 billion payout in authors’ lawsuit over Anthropic’s alleged use of pirated books to train AI

A federal judge in San Francisco granted preliminary approval to a $1.5 billion settlement resolving claims that Anthropic used hundreds of thousands of pirated books to train its AI systems. The deal, which would pay about $3,000 per title for roughly 465,000 works, drew praise from publishing groups and the Authors Guild, who called it a milestone in holding AI developers accountable. Judge William Alsup, who previously found AI training to be transformative fair use while faulting Anthropic’s acquisition of materials from pirate sites, cautioned that administering payments will be complex and pressed for robust notice to ensure broad participation. Anthropic welcomed the approval, saying it preserves the court’s earlier fair-use ruling and narrows the dispute to sourcing practices. The case underscores growing legal pressure on AI firms to license content, even as courts sort out the boundaries of fair use in training data. Alsup also said he plans to step down from the bench by year-end.

Read more