- A federal judge granted final approval on July 20, 2026 to Anthropic’s $1.5 billion settlement with authors and publishers, the largest known payout in U.S. copyright history.
- The deal pays roughly $3,000 per work across more than 480,000 books that Anthropic downloaded from pirate libraries to train its Claude models.
- The settlement resolves how Anthropic obtained the books, but leaves standing an earlier ruling that training AI on copyrighted text can be fair use.
- Parallel copyright lawsuits against Google, Meta, Midjourney, and OpenAI remain unresolved, so the industry-wide question is still open.
What Happened
U.S. District Judge Araceli Martinez-Olguin granted final approval on Monday, July 20, 2026, to Anthropic’s $1.5 billion settlement of a class action copyright lawsuit brought by a group of authors and book publishers, TechCrunch reported, citing Reuters. The ruling clears Anthropic to begin paying out claimants who accused the company of misusing their books to train its Claude chatbot. Judge Martinez-Olguin inherited the case from Judge William Alsup of the Northern District of California, who issued preliminary approval last year and has since retired. Several authors objected that the payout was too small, but the judge rejected those arguments, writing that the complaints were “not grounded in a realistic assessment of the overall risks and rewards of a trial.”
Why It Matters
The approval finalizes the largest known settlement in the history of U.S. copyright law, yet it settles far less than its size suggests. Alsup had already sided with Anthropic on the central question, ruling that training an AI model on copyrighted text can qualify as fair use — a decision widely read as a turning point for how AI companies source training data. His ruling drew a hard line, though: buying and scanning books was permissible, but downloading them from pirate sites was not, and that piracy question could have gone to a jury. Anthropic agreed to settle rather than face trial and potential statutory damages. Because the case ends here and was decided by a single district court, it never reaches an appeals court and creates no binding precedent, leaving other judges free to reach their own conclusions.
Technical Details
The money breaks down to roughly $3,000 per work across more than 480,000 covered books. Anthropic assembled its training library from two sources: titles it purchased and scanned, which Alsup found lawful, and books it pulled from pirate repositories including Library Genesis and Pirate Library Mirror, which he found illegal on their own terms. Reuters reported that Anthropic saved more than 7 million pirated books to what the court described as a “central library.” The settlement also resolved a fight over legal fees: the judge awarded class counsel $101 million of the $187.5 million they had requested. The distinction the court drew — lawful use of lawfully obtained text versus liability for pirated acquisition — is the practical takeaway other AI labs are now studying, a question that also runs through disputes over how model outputs and datasets are sourced.
Who’s Affected
The immediate beneficiaries are the authors and publishers who hold rights to the roughly 480,000 works. Anthropic deputy general counsel Aparna Sridhar said “more than 91% of authors and publishers covered by the settlement have claimed their share,” and framed the outcome around the underlying legal finding: the company reached the deal “after the court’s landmark ruling that training AI on books is fair use under copyright law — which remains the law today.” Not everyone views the result as a win. A subset of rights holders objected that the sum was too low, that attorney fees were excessive, and that some copyright owners were wrongly excluded; the judge overruled each objection. Some authors and publishers opted out entirely and are continuing separate litigation against Anthropic, which has publicly pushed for clearer rules governing AI development even as it fights these claims.
What’s Next
The broader legal fight over AI training data now moves to other courtrooms. The Anthropic case closes one front, but a string of copyright suits remains active against Google, Meta, Midjourney, and OpenAI over whether training on copyrighted works without permission is lawful. Just last week, a group of publishers and authors — including Hachette, Cengage, Elsevier, and novelist Scott Turow, alongside the writers’ group S.C.R.I.B.E. — filed a class action against Google, accusing it of using their copyrighted works to train its Gemini platform. With Anthropic’s payments set to begin flowing to claimants, the settlement gives the industry a concrete price tag for pirated training data, even as the question of what counts as fair use plays out case by case in courts that are not bound by the outcome here.
