A federal judge in San Francisco approved Anthropic’s $1.5 billion settlement in a class-action lawsuit brought by authors who accused the artificial intelligence maker of storing millions of pirated books to train its Claude chatbot. The payout represents the largest copyright settlement in U.S. history.

The settlement resolves claims that Anthropic saved more than 7 million stolen book copies into an internal library, a practice the court found to be infringement separate from training the model. While the same judge previously ruled that using books for AI training qualified as fair use, the company faced trial risks that could have pushed damages into the hundreds of billions. The agreement covers thousands of writers and nearly 500,000 works, averaging roughly $4,000 per title, though some authors opted out or criticized the fee structure.