A federal judge has approved Anthropic’s $1.5 billion class action settlement with authors who accused the AI company of using copyrighted books to train its models without permission, according to Reuters and a Monday court order from Judge Araceli Martínez-Olguín.
The approval gives authors and publishers covered by the deal a route to compensation of about $3,000 for each book allegedly pirated by Anthropic. For writers whose work ended up in training datasets through unauthorized downloads, the order converts one of the AI industry’s messiest copyright fights into a very expensive cleanup bill.
Judge Martínez-Olguín wrote in the order that the settlement provides “meaningful relief.” The law firm representing the plaintiffs, Susman Godfrey, called the $1.5 billion deal the “largest known copyright recovery in history.” That is the plaintiffs’ framing, but the number is large by any copyright-litigation standard.
The case began in 2024, when authors Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson sued Anthropic. Their claims targeted the company’s use of books in AI training, a central dispute in the current wave of copyright cases against AI developers. Training a large language model requires enormous text datasets. The fight here was not only over whether copyrighted books can be used for training, but over how Anthropic allegedly got them.
Anthropic previously won part of the case before Judge William Alsup, who has since retired. Alsup gave the company a partial victory last year, but he also allowed a separate class action to move forward over allegations that Anthropic engaged in what he described as “Napster-style downloading of millions of works.”
That distinction matters. AI companies have often argued that training on copyrighted material can qualify as fair use. The downloading allegations put a different problem on the table: whether the company acquired copies of books through piracy before feeding them into its systems. A fair-use argument does not magically launder the method of acquisition, which is why this case stayed dangerous for Anthropic even after its partial win.
Anthropic agreed to the $1.5 billion settlement in September 2025. Alsup later gave it preliminary approval, according to Reuters. Martínez-Olguín’s order now grants final approval.
Anthropic general counsel Aparna Sridhar told Reuters that the company is pleased that more than 91 percent of covered authors and publishers have claimed their share. Sridhar said Anthropic is looking forward to bringing the matter to a close.
The settlement does not end Anthropic’s copyright exposure. Reuters has reported that Chicken Soup for the Soul has sued the company, and Publishers Weekly has reported that several other authors have filed claims arguing that $3,000 per book is not enough. So Anthropic has settled one major book case, but the broader question facing AI companies remains unresolved: who gets paid when copyrighted works are used to build commercial models, and how much.
This story draws on original reporting from The Verge.
