Anthropic to pay authors $1.5B to settle lawsuit over pirated chatbot training material
Anthropic agreed to a proposed $1.5 billion settlement with authors over pirated books used for chatbot training, subject to court approval.
Summary
Michigan Public, carrying an NPR report updated on September 5, 2025, says Anthropic will pay 1.5 billion dollars to settle a class-action suit from book authors. The writers claim the company obtained illicit copies of their books and used them for Claude's training. A judge could approve the deal as early as Monday, and U.S. District Judge William Alsup has set a hearing that day to review the terms. The authors' lawyer, Justin Nelson, called it the largest copyright recovery ever and the first of that type in the AI age.
Payment works out to roughly 3,000 dollars per title, across an estimated 500,000 books. Authors Andrea Bartz, Charles Graeber and Kirk Wallace Johnson sued the previous year and represent a wider group of writers and publishers. The settlement further obliges Anthropic to delete the original files of the books it downloaded.
The background is a June ruling by Alsup that training chatbots on copyrighted books was not illegal, treating it as fair use, but that Anthropic wrongfully obtained millions of books from pirate sites. He found more than 7 million pirated books, drawn from Books3, Library Genesis and the Pirate Library Mirror. Anthropic later bought books in bulk and scanned them, which the judge said did not erase the earlier piracy. Analyst William Long said a trial loss in December could have cost multiple billions of dollars.
Reactions varied. The Authors Guild called it an excellent result, while the Danish Rights Alliance said it would do little for European authors whose works are not registered with the U.S. Copyright Office. Long expects it to influence other disputes, such as the suit against OpenAI and Microsoft.