US District Judge Araceli Martinez-Olguin approved Anthropic's $1.5 billion settlement with a group of authors who sued the AI company for copyright infringement in training its Claude models.

The lawsuit, filed in 2024, accused Anthropic of using pirated copies of books to train Claude without permission. The case covered more than 480,000 works stored in what the company called a central library.

Judge William Alsup, who originally presided over the case, had ruled that Anthropic's use of the writers' work for training counted as fair use. However, he found the company infringed copyright by storing 7 million pirated books in its central repository.

Alsup initially rejected the settlement deal in September, calling it incomplete and saying it was "shoving the terms down the throat of authors." He retired before the revised agreement was finalized.

Settlement terms and payouts

Under the approved settlement, authors will receive approximately $3,000 per affected work. Anthropic must also destroy all pirated copies of the books used in training.

The company launched an official settlement website listing affected works so authors could determine their eligibility for claims. The settlement represents one of the largest copyright payouts in AI training disputes to date.

Some plaintiffs opted out of the settlement, finding the $3,000-per-work payout insufficient. These authors plan to pursue separate lawsuits against Anthropic for potentially higher damages.

The case highlights ongoing tensions between AI companies and content creators over training data usage. Similar lawsuits against other AI firms remain pending in federal courts.