Anthropic's $1.5 Billion Settlement Over Pirated Books in AI Training Gets Final Approval
The approval sets a record as the largest settlement in U.S. copyright litigation history, but the opt-outs and separate lawsuits mean the legal question of using pirated data for AI training is far from settled.
Key Facts
- The settlement covers over 480,000 works used by Anthropic to train its AI model Claude.
- Anthropic agreed to pay a total of $1.5 billion in settlement funds, paid in four installments.
- Writers are expected to receive approximately $3,000 per eligible work.
- Anthropic is required to delete the pirated book data it held.
- Claims have been filed for over 440,000 works, more than 90% of the covered total.
Reporting from 2 sources: GameBusiness.jp, GIGAZINE.
On July 20, 2026, a U.S. federal court in San Francisco formally approved a $1.5 billion settlement between AI developer Anthropic and a group of writers who sued the company in 2024. The lawsuit alleged that Anthropic used hundreds of thousands of copyrighted books from pirate sites to train its AI model Claude. The case, Bartz v. Anthropic PBC, was initially handled by Judge William Alsup, who ruled that while training AI on legally acquired books was fair use, downloading over 7 million books from pirate sites to build a central library was not. Alsup rejected an earlier settlement proposal in September 2025, citing issues with the list of covered works and notification procedures. After Anthropic launched a settlement website and published an eligible works list, Judge Araceli Martinez-Olguin, who succeeded the retired Alsup, approved the terms. The settlement covers over 480,000 works, with writers receiving approximately $3,000 per eligible work. Anthropic must also delete the pirated book data. Claims have been filed for over 440,000 works, more than 90% of the covered total. However, some authors and publishers have opted out of the settlement and filed separate lawsuits against Anthropic, which remain pending.
Judge Araceli Martinez-Olguin of the U.S. District Court for the Northern District of California formally approved the settlement on July 20, 2026, after Judge William Alsup, who had initially handled the case, retired. Alsup had rejected an earlier settlement proposal in September 2025, citing issues with the list of covered works and notification procedures, and saying the terms were "forced on the writers."
Some authors opposed the settlement on the grounds that "the settlement amount is insufficient," "the plaintiffs' attorneys' fees are excessive," and "some copyright owners are improperly excluded." These plaintiffs have opted out of the class settlement and filed individual lawsuits against Anthropic, which remain pending.
According to the plaintiffs' legal team, claims have been filed for over 440,000 works, more than 90% of the covered total. The team noted that this rate far exceeds the typical claim rate of about 10% in class action lawsuits. Anthropic will contribute the $1.5 billion settlement fund in four installments. The company must also delete the pirated book data it held.
Synthesized by Yomimono from the 2 cited sources below, including Japanese-language reporting where cited, then editorially reviewed before publishing.