A $1.5 billion settlement meant to resolve claims that Anthropic trained its AI models on pirated books is facing fresh scrutiny in federal court. U.S. District Judge William Alsup called the judgment “full of pitfalls,” signaling that the high-profile copyright case is far from over and could reshape how AI companies handle training data.
The current deal would pay roughly $3,000 for each of about 465,000 book titles allegedly used without permission to train the company’s Claude models. Concerned that the number could balloon with new claims “coming out of the woodwork,” the judge ordered a “drop-dead list” by September 15 that definitively identifies the works at issue. He also set another hearing for September 25 to revisit the terms and logistics of the settlement.
Leaders in the publishing world pushed back on the court’s approach. Maria Pallante of the Association of American Publishers said the ruling reflected “a lack of understanding of how the publishing industry works,” emphasizing that the list of covered works must be complete and that the court’s reluctance to allow more time—without explanation—was troubling. She also warned that the court appeared to be envisioning an “administratively challenging” claims process that could be unworkable for authors and publishers and trigger years of collateral litigation.
Authors Guild CEO Mary Rasenberger said she was “shocked by the court’s offhand suggestion,” voicing concern that the process could end up pressuring authors rather than supporting them. According to the Guild, their role should be to inform and assist rights holders, not to drive them into an onerous or confusing claims maze.
Anthropic, backed by major tech investors, was sued last year by authors who say their books were downloaded and used to train Claude without permission. The proposed payout is being described as the largest publicly reported recovery in the history of U.S. copyright litigation, underscoring the stakes for both the AI industry and the creative community.
What happens next could set an important precedent. The court wants certainty about the scope of the alleged infringement and a practical, fair way for authors to claim compensation. With a critical hearing looming, the outcome may shape the future of AI training practices, the boundaries of fair use, and the safeguards required to protect copyrighted works in the age of large-scale machine learning.
