NEW YORK CITY, New York: Artificial intelligence company Anthropic has agreed to pay US$1.5 billion to settle a class-action lawsuit filed by authors who said the company used pirated copies of their books to train its chatbot Claude.
The agreement, disclosed late last week, would resolve claims brought last year by thriller novelist Andrea Bartz and nonfiction writers Charles Graeber and Kirk Wallace Johnson. The authors alleged that Anthropic downloaded millions of copyrighted works without permission from online shadow libraries. The case later expanded to include hundreds of thousands of authors and publishers.
Under the terms of the settlement, affected authors or publishers will receive about $3,000 per book for an estimated 500,000 works included in the company’s training data. Anthropic has also agreed to delete the book files it obtained.
A hearing to review the settlement is scheduled for Monday before U.S. District Judge William Alsup in San Francisco.
In June, Alsup issued a mixed ruling in the case. He found that using copyrighted books to train an AI model did not in itself violate copyright law, but said Anthropic had wrongfully acquired more than seven million books from pirate websites. Those included nearly 200,000 titles from the Books3 dataset, five million from Library Genesis, and two million from the Pirate Library Mirror.
Anthropic said in a statement that the settlement, if approved, would “resolve the plaintiffs’ remaining legacy claims.” Aparna Sridhar, the company’s deputy general counsel, said Anthropic remains focused on developing “safe AI systems that help people and organizations extend their capabilities, advance scientific discovery, and solve complex problems.”
The Authors Guild, which represents thousands of U.S. writers, welcomed the agreement. CEO Mary Rasenberger said it was “an excellent result for authors, publishers, and rightsholders generally,” and warned that the case showed “serious consequences when [AI companies] pirate authors’ works to train their AI.”
The group had previously estimated damages could be as low as $750 per work if the case went to trial. The higher $3,000-per-book award likely reflects a smaller pool of works, excluding duplicates or books without copyright protection.
The Danish Rights Alliance said the settlement may not help European writers and publishers whose works were also copied but are not registered with the U.S. Copyright Office.
Anthropic, a San Francisco-based company founded in 2021 by former OpenAI executives, is one of the largest competitors in the AI industry. This week, it reported that it had raised $13 billion in new funding and valued itself at $183 billion. It expects about $5 billion in sales this year but has not reported a profit.
The settlement could influence other copyright disputes against AI firms. OpenAI and its partner Microsoft are facing lawsuits from authors and newspapers. Meta and Midjourney are also defendants in similar cases. On September 5, a separate group of authors filed a new copyright suit against Apple in the same federal court.