Judge skewers $1.5B Anthropic settlement with authors in pirated books case over AI training

In a significant development in the ongoing legal battle between Anthropic, an artificial intelligence company, and a group of authors, a federal judge has expressed serious concerns regarding a proposed $1.5 billion settlement. The settlement is aimed at resolving allegations that Anthropic illegally used nearly 500,000 copyrighted books to train its chatbots, raising the possibility of a trial still being on the horizon. During a recent hearing in San Francisco, U.S. District Judge William Alsup spent considerable time criticizing the settlement agreement, highlighting potential issues he believes could complicate its approval. "We'll see if I can hold my nose and approve it," Alsup remarked, indicating his reservations about the deal. A follow-up hearing has been scheduled for September 25 to further explore these concerns. The judge’s skepticism comes just days after Anthropic announced the settlement alongside the attorneys representing the authors. The deal aims to preempt a trial that was set to commence in December. Back in June, Judge Alsup delivered a mixed ruling on the case, affirming that while it isn’t illegal to train AI models on copyrighted materials, Anthropic had indeed wrongfully acquired a significant number of books from pirated sources to enhance its Claude chatbot. If approved, the settlement would offer approximately $3,000 to each author and publisher for the books included in the agreement. According to Justin Nelson, the attorney representing the authors, around 465,000 titles are identified as having been pirated by Anthropic. However, Alsup expressed the need for more definitive assurances that this number wouldn’t increase, fearing that additional lawsuits could emerge. One of the judge's primary concerns revolved around the claims process, emphasizing the importance of ensuring that all eligible authors are informed and able to participate. He requested a comprehensive list of the pirated books by September 15 and a claims form by September 22 for his review prior to the next hearing. Additionally, Alsup raised alarms about the involvement of two major organizations in the publishing industry—the Authors Guild and the Association of American Publishers—suggesting they might exert pressure on authors to accept the settlement without fully grasping its implications. Both organizations had representatives present at the hearing but did not speak. Among those in attendance were authors Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson, who have been vocal about their opposition to the unauthorized use of their works. Johnson described the settlement as a starting point for a broader struggle against the unchecked advancement of AI at the expense of human creativity. Despite the attorney’s assurances that the settlement would be handled transparently, Judge Alsup’s apprehensions suggest that the path forward remains uncertain, and he may ultimately decide to let the case proceed to trial.