Artificial intelligence startup Anthropic has agreed to pay a record-setting $1.5 billion to a group of book authors and publishers in order to settle a class action lawsuit. The payout is thought to be the largest in the history of U.S. copyright suits and could influence other cases where an AI company has been sued for copyright violations.
Most Read from Fast Company
“This settlement sends a powerful message to AI companies and creators alike that taking copyrighted works from these pirate websites is wrong,” Justin Nelson, an attorney for the plaintiffs, said in a statement.
The suit, filed last year, was brought by authors Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson over copyright infringement. They alleged that Anthropic used the authors’ copyrighted books to train its chatbot, Claude.
In June, a judge ruled that while Anthropic was allowed to train its AI model using books that it had acquired the copyright for under fair use rules, the startup had illegally acquired books via online libraries that contained bootleg copies of books. The judge concluded that the authors had cause for the case to proceed to a trial. That was slated to start in December.
Now, according to a court filing on Friday, the startup has agreed to pay authors $3,000 for each of around 500,000 books it used to train its AI, plus interest. The company also agreed to destroy the datasets containing the allegedly pirated material.
“As best as we can tell, it’s the largest copyright recovery ever,” Nelsons said, according to the Associated Press “It is the first of its kind in the AI era,” he added.
“Today’s settlement, if approved, will resolve the plaintiffs’ remaining legacy claims. We remain committed to developing safe AI systems that help people and organizations extend their capabilities, advance scientific discovery, and solve complex problems,” Anthropic’s deputy general counsel Aparna Sridhar said in a statement.
The settlement may give other AI companies facing similar challenges to their use of copyrighted material to train their models pause. Back in June, the judge in this case affirmed that using books to train a large language model represented a “transformative” use of the work—crucial for “fair use”—but made a clear distinction if the books had been sourced illegally.
