started · updated
Anthropic agrees to $1.5 billion copyright settlement with authors
A U.S. federal judge in San Francisco approved Anthropic’s $1.5 billion settlement with a class of authors and publishers who alleged the company used pirated books to train its Claude chatbot. The agreement, signed by Judge Araceli Martínez‑Olguín after an earlier preliminary approval by Judge William Alsup in September 2025, provides roughly $3,000 per work for about 500,000 titles, covering more than 91 % of the claims filed. Anthropic will also pay around $101 million in attorney fees. The case, the largest copyright class action in U.S. history, centered on the unlawful acquisition of millions of books from sites such as LibGen and Pirate Library Mirror, not on the fairness of using the texts for AI training. The court found that while training on legally obtained works may fall under fair‑use, the massive download and storage of pirated copies was not protected. The settlement ends the financial exposure for Anthropic but does not resolve the broader legal question of whether copyrighted material can be used to train large‑language models without permission. Some authors have opted out of the deal to pursue separate lawsuits, and the issue remains active in other AI‑related copyright suits.