Judge approves $1.5B Anthropic settlement for pirated books used to train Claude

A federal judge has approved a $1.5 billion settlement between Anthropic and a group of authors and publishers who alleged the company trained its Claude models on pirated copies of their books. The case, captioned Bartz v. Anthropic, was reported July 21 by the Associated Press and circulated further via Hacker News.

Details beyond the settlement figure and the underlying allegation, that copyrighted works were used without authorization to build Claude’s training data, were not specified in the available reporting. The approval represents a court-sanctioned resolution rather than a trial verdict, closing the matter under terms the judge found acceptable.

The ruling arrives as one of the first major monetary settlements tied directly to the practice of training large language models on copyrighted text obtained without license. Publishers and authors’ groups have pressed similar claims against other AI developers, and this outcome, whatever precedent it sets, will likely be cited in those ongoing disputes over what training data may lawfully be used and what compensation is owed when it is not.

No further particulars, such as distribution mechanisms for the settlement fund, admissions of liability, or effects on Claude’s continued availability, were included in the source material reviewed for this bulletin. The observatory notes the date, the parties, and the sum, and defers further characterization until additional documentation becomes available.