A federal judge in San Francisco approved Anthropic’s $1.5 billion settlement with authors who accused the company of training its Claude models on pirated books, closing what lawyers for both sides describe as the largest copyright recovery in U.S. history.
Judge Araceli Martínez-Olguín got final approval on July 20, months after she objected to the settlement and pressured plaintiffs’ attorneys to give her a clearer accounting of their fees.
money is resolved Bartz vs. Anthropica class action lawsuit filed in 2024 by novelists Andrea Bartz and Charles Graeber and nonfiction writer Kirk Wallace Johnson, on behalf of approximately half a million works.
At about $3,000 per book, the payment dwarfs the going rate in most cases. AI Copyright Disputesand lands as comparable cases against Meta, Google and OpenAI continue.
The 💜 of EU technology
The latest rumors from the EU tech scene, a story from our wise founder Boris and some questionable AI art. It’s free, every week, in your inbox. Register now!
The case generated an unusual division. In a June 2025 ruling, Judge William Alsup determined that training a model with legally purchased books was fair use, a first-of-its-kind victory for the industry, but that Anthropic’s decision to download millions of titles from shadow libraries LibGen and Pirate Library Mirror was not.
That finding left the company exposed to legal damages large enough to threaten its bottom line, which could run into hundreds of billions of dollars if the case had gone the wrong way.
Anthropic agreed to settle in September 2025 rather than put the piracy issue before a jury, and because this is a settlement and not a verdict, it does not establish any binding precedent, a distinction the company has wanted to keep in view.
“We reached this agreement in 2025, after the court’s landmark ruling that training AI on books is fair use under copyright law, which remains the law today.” said Aparna Sridhar, deputy general counsel at Anthropic.
Martínez-Olguín, who inherited Alsup’s approval stage, seemed less interested in the fair use headline than in whether the deal was fair to the authors it sought to compensate.
“The $1.5 billion settlement provides substantial benefits to the class in light of the novel claims made,” she wrote, adding that “Success at trial was not assured and a loss would have left the class without recourse.”
Their previous hesitation had focused mainly on who gets paid and how much. He wanted a more complete estimate of the class’s attorneys’ fees, a multimillion-dollar expense reserve and service awards set aside for the three plaintiffs named before he signed the order.
In granting approval, he scaled back the attorneys’ request, cutting the attorneys’ fees by about $86 million from what the group’s attorneys had requested, according to Law.com.
The plaintiffs’ own framework had been considerably less measured than the judge’s, and in their complaint they accused Anthropic of having “attempted to steal the fire of Prometheus” by “exploiting the human expression and ingenuity behind each of those works.”
For the authors, the remaining questions largely have to do with paperwork. Rightsholders whose titles appear in the list of works pulled from LibGen and PiLiMi were asked to submit claims earlier this year, and the standard commercial division sends half the payment for each book to the publisher and the other half to the author, unless a contract says otherwise, according to the Authors Guild.
With final approval now in place, disbursement to eligible rights holders can begin.
The industry in general will read the number more than the reasoning. A $1.5 billion sum concentrates minds in an industry that has treated large-scale scraping as a cost of doing business, and comes as editors pursue Goal and Newspapers sue OpenAI about the same underlying practice.
None of those cases have yet produced Alsup’s clear distinction between legal training and illegal acquisition, meaning the next defendant to face a jury may not have the opportunity to settle on such content terms.
Anthropic, for its part, keeps the fair use rule it wanted and pays for the library it should never have downloaded.
