A federal judge in San Francisco has granted final approval to a $1.5 billion settlement between the artificial intelligence company Anthropic and a class of authors, closing the largest known copyright damages case in United States history. On the same day, the University of Tennessee Research Foundation filed a patent infringement lawsuit against Anthropic in Delaware, alleging that the firm’s AI systems violate its patents on neuroscience-inspired machine learning technology. The twin developments underscore the intensifying legal pressure on AI developers over the data and methods used to build large language models.

The settlement resolves a 2024 class action in which authors accused Anthropic of using pirated copies of their books to train the chatbot Claude. In a June 2025 ruling, Judge William Alsup held that training AI on books constitutes fair use under US copyright law, but found that Anthropic’s storage of more than seven million pirated books in a central library—not all of which were used for training—violated copyright. The company’s deputy general counsel said the settlement was reached after that landmark fair-use decision. The lead attorney for the authors called the payout “the largest copyright recovery settlement in history.” Some authors objected that the sum was too low, but Judge Araceli Martínez-Olguín overruled those objections, stating they were not based on a realistic assessment of the risks of a trial.

The settlement covers roughly 500,000 rights holders, yielding an average of about $3,000 per book. A number of authors and publishers opted out and have filed separate lawsuits that are still pending. The newly launched patent suit alleges that Anthropic’s systems infringe two patents covering machine-learning technology inspired by neuroscience, invented by university professors. The university is seeking unspecified damages and an injunction, and stated that Anthropic’s “disregard for third-party intellectual property rights extends beyond the use of copyrighted material.”

Court filings have revealed the lengths to which Anthropic went to obtain training data. According to documents cited in the copyright case, the company downloaded millions of books from a digital archive that openly flouted copyright law. After the court’s fair-use ruling, Anthropic hired Tom Turvey, the former Google executive who built Google Books, to acquire physical copies of books from retailers, remove their spines, scan the pages and then shred the volumes. Judge Alsup noted that millions of books had already been turned into pulp. In Moscow, media companies, IT developers and rights-holder associations have petitioned the government to insert licensing requirements into a draft law on AI development. The current text would permit training on any publicly available copyrighted work without consent, a provision the petitioners argue would legalise the uncompensated use of all internet content.

The settlement distribution is expected to begin shortly. The patent case is in its preliminary stages, and the opt-out copyright lawsuits continue. Legislative debates in Russia and elsewhere signal that the legal framework governing AI training data remains deeply contested.