Court grants final approval to $1.5bn Anthropic book-piracy settlement

A federal judge in the Northern District of California granted final approval to Anthropic’s settlement of Bartz v. Anthropic, a class action brought by authors including Andrea Bartz, Charles Graeber and Kirk Wallace Johnson over the company’s use of pirated books to train its models. The $1.5 billion settlement, first agreed roughly ten and a half months earlier, was described by the Authors Guild as the largest class-action copyright recovery in US history. Under its terms, authors receive approximately $3,000 per qualifying work — about four times the $750 statutory minimum for copyright infringement — with contact information successfully reached for authors covering 99.5% of the roughly 595,000 works on the settlement’s list.

The settlement addressed only the narrower half of the underlying case. Judge William Alsup had ruled in 2025 that training Anthropic’s models on books it had legitimately purchased was fair use, but that downloading pirated copies from sites including Library Genesis and the Pirate Library Mirror was not; the settlement resolved liability for that piracy through August 2025, while leaving authors free to pursue any claims involving works not on the settlement’s list. Final approval followed 54 objections, which the court addressed on their merits before approving the deal; requested attorneys’ fees were reduced to roughly $101.6 million, about 6.8% of the total settlement, and service awards to the three named plaintiffs were trimmed to $15,000 each. The court also ordered Anthropic to destroy the pirated files it had acquired.

The ruling closed out the most advanced of the major AI copyright cases to reach a monetary resolution, and gave other plaintiffs suing AI companies over training data a concrete number — cost per pirated work — to point to in settlement negotiations elsewhere, even as the fair-use question for purchased material remained good law only within this case’s jurisdiction.