A federal judge based in San Francisco has approved Anthropic's $1.5 billion agreement with writers who claimed the firm trained its Claude systems using unauthorized book copies, marking the largest copyright recovery in United States history as described by lawyers from both sides.

Judge Araceli Martínez-Olguín issued final approval on 20 July after earlier raising questions about the proposed legal fees and seeking more details from the plaintiffs' representatives.

The resolution ends Bartz v. Anthropic, a class action filed in 2024 by novelists Andrea Bartz and Charles Graeber along with nonfiction author Kirk Wallace Johnson, covering roughly half a million titles.

At roughly $3,000 per book, the compensation exceeds standard rates in most AI copyright matters and comes as similar actions against Meta, Google, and OpenAI continue without resolution.

The matter featured a notable distinction. In a June 2025 decision, Judge William Alsup ruled that training on legally obtained books qualified as fair use in a first ruling of its kind for the sector, yet acquiring millions of volumes from unauthorized sources such as LibGen and Pirate Library Mirror did not qualify.

This left the company facing potential statutory damages that could have reached hundreds of billions and threatened its financial position if the case had proceeded unfavorably.

Anthropic chose to settle in September 2025 instead of presenting the piracy issue to a jury. Because the outcome is a settlement rather than a verdict, it creates no binding precedent, a point the company has stressed.

"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 took over the approval process from Alsup, emphasized fairness to the authors over the fair-use aspect. "The $1.5 billion settlement provides substantial benefits to the class in light of the novel claims asserted," she wrote, noting that "success at trial was not assured, and a loss would have left the class with no recourse."

Her earlier concerns focused mainly on payment allocation, including a detailed review of class counsel fees, a reserve for expenses, and service awards for the three named plaintiffs.

In granting approval she reduced the attorneys' fee request by roughly $86 million from the amount originally sought by class counsel, according to Law.com.

The plaintiffs had framed their case more strongly, accusing Anthropic of attempting to take creative works by extracting expression and ingenuity from each title.

For writers, remaining steps involve paperwork. Rightsholders with titles on the list from LibGen and PiLiMi were required to submit claims earlier this year, with payments typically divided equally between publisher and author unless contracts specify otherwise, per the Authors Guild.

With approval secured, payments to eligible rightsholders can now start.

The broader sector will focus on the total amount. The $1.5 billion figure draws attention in an industry that has viewed large-scale data collection as routine, arriving as publishers target Meta and newspapers pursue OpenAI over similar practices.

Those matters lack the clear separation between lawful training and unlawful acquisition seen here, so the next defendant facing a jury may not have the option to settle on comparable terms.

Anthropic secures the fair-use ruling it sought while covering the cost of the library it should not have accessed.