A proposed $1.5 billion settlement between artificial intelligence company Anthropic and a coalition of authors has run into trouble after a federal judge questioned the deal’s fairness and design.
US District Judge William Alsup in San Francisco sharply criticized the agreement on Monday, saying it was incomplete and left key issues unresolved.
The proposed settlement aimed to resolve allegations that Anthropic used roughly 500,000 pirated books to train its Claude chatbot. The plan would give authors and publishers about $3,000 per title.
However, Judge Alsup balked at the details. “I have an uneasy feeling about all the hangers on in the shadows,” he said in court, according to the Associated Press. He added, “We’ll see if I can hold my nose and approve it” at a future hearing.
Bloomberg also reported that Alsup also said he felt “misled” by the way the agreement was presented, warning that the deal seemed to be pushed “down the throat of authors.”
Deadlines and demands
Alsup ordered the parties to produce by Sept. 15 a complete list of the disputed books, estimated at 465,000, and to provide him with a sample claims form no later than Sept. 22. He scheduled a follow-up hearing on Sept. 25 to evaluate the progress.
He stressed that the process must guarantee proper notification for every author eligible to claim compensation. Without clear rules, he warned, authors risk being excluded and “get[ting] the shaft.”
Reactions from authors and publishing groups
Justin Nelson, attorney for the authors, told the court that he and his team “care deeply that every single proper claim gets compensation.” He argued that the case has received significant publicity, ensuring authors are aware of the settlement. “This is not an under-the-radar warranty case,” Nelson said, as quoted by the Associated Press.
Outside the courtroom, author Kirk Wallace Johnson, one of the plaintiffs, framed the fight in larger terms. He described the settlement as the “beginning of a fight on behalf of humans that don’t believe we have to sacrifice everything on the altar of AI,” according to the Associated Press.
However, major industry groups pushed back against the judge’s skepticism. Maria Pallante, CEO of the Association of American Publishers, said in a statement that the court “demonstrated a lack of understanding of how the publishing industry works” and warned that the judge’s vision of the claims process “would be unworkable, and sees a world with collateral litigation between authors and publishers for years to come,” the Associated Press reported.
Authors Guild CEO Mary Rasenberger also defended her group’s role, rejecting the notion that it was working behind the scenes to pressure authors. She said that the Guild was “confused” by the suggestion it was pressuring authors and insisted the group’s role was to “ensure that authors’ interests are fully represented” and that it would not benefit financially from the deal.
The case, Bartz v. Anthropic PBC, had been set for a December trial before the settlement surfaced earlier this month. With Alsup delaying approval, the outcome is uncertain.
If the settlement cannot be finalized by the September deadlines, the case is expected to head to trial — a proceeding that could shape how courts handle copyrights disputes involving AI firms such as OpenAI, Meta, and others.
Anthropic’s deal in a high-profile copyright case signals more than closure — it sets the stage for how generative AI firms may navigate future legal battles. See why this settlement could reshape the rules for the industry.