A federal judge in San Francisco has signed off on Anthropic’s $1.5 billion settlement with authors who accused the company of training its Claude models on pirated books, closing what lawyers on both sides describe as the largest copyright recovery in United States history.
Judge Araceli Martínez-Olguín granted final approval on 20 July, months after she balked at the deal and pressed the plaintiffs’ lawyers for a clearer account of their fees.
The money resolves Bartz v. Anthropic, a class action brought in 2024 by the novelists Andrea Bartz and Charles Graeber and the non-fiction writer Kirk Wallace Johnson, on behalf of roughly half a million works.
At about $3,000 per book the payout dwarfs the going rate in most AI copyright disputes, and it lands while comparable cases against Meta, Google, and OpenAI grind on.
The case turned on an unusual split. In a June 2025 ruling, Judge William Alsup found that training a model on lawfully purchased books was fair use, a first-of-its-kind win for the industry, but that Anthropic’s decision to download millions of titles from the shadow libraries LibGen and Pirate Library Mirror was not.
That finding left the company exposed to statutory damages large enough to threaten its balance sheet, potentially running to hundreds of billions of dollars had the case gone the wrong way.
Anthropic agreed to settle in September 2025 rather than put the piracy question to a jury, and because this is a settlement and not a verdict, it establishes no binding precedent, a distinction the company has been keen to keep in view.
“We reached this settlement 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, Anthropic’s deputy general counsel.
Martínez-Olguín, who inherited the approval stage from Alsup, seemed less taken with the fair-use headline than with whether the deal was fair to the authors it claimed to compensate.
“The $1.5 billion settlement provides substantial benefits to the class in light of the novel claims asserted,” she wrote, adding that* “success at trial was not assured, and a loss would have left the class with no recourse.”*
Her earlier hesitation had been mostly about who gets paid, and how much. She wanted a fuller reckoning of class counsel fees, a multimillion-dollar expense reserve, and the service awards set aside for the three named plaintiffs before she would put her signature to the order.
In granting approval she trimmed the lawyers’ request, cutting the attorneys’ fee award by roughly $86 million from what class counsel had sought, according to Law.com.
The plaintiffs’ own framing had been rather less measured than the judge’s, their complaint accusing Anthropic of having “attempted to steal the fire of Prometheus” by “strip-mining the human expression and ingenuity behind each one of those works.”
For authors, the remaining questions are largely about paperwork. Rightsholders whose titles appear on the works list drawn from LibGen and PiLiMi were asked to file claims earlier this year, with the standard trade split sending half of each book’s payment to the publisher and half to the author unless a contract says otherwise, according to the Authors Guild. With final approval now in hand, disbursement to eligible rightsholders can begin.
The wider industry will read the number rather than the reasoning. A sum of $1.5 billion concentrates minds in a sector that has treated large-scale scraping as a cost of doing business, and it arrives as publishers pursue Meta and newspapers sue OpenAI over the same underlying practice.
None of those cases has yet produced Alsup’s tidy distinction between lawful training and unlawful acquisition, which means the next defendant to face a jury may not get the chance to settle on such contained terms.
Anthropic, for its part, keeps the fair-use ruling it wanted and pays for the library it should never have downloaded.
Get the TNW newsletter #
Get the most important tech news in your inbox each week.