A Judge Approves Anthropic’s $1.5 Billion AI-Copyright Settlement
A federal judge approved Anthropic's $1.5 billion settlement with authors who alleged the company used pirated books to train Claude, creating a landmark AI-copyright payout.
A federal judge has approved Anthropic’s $1.5 billion settlement with authors who alleged that the AI company used pirated books to train Claude. The approved agreement will pay thousands of authors about $3,000 per book, according to The Associated Press, making it one of the largest payouts yet in the still-forming legal battle over AI training data.
The case concerns a question that has followed generative AI from its earliest successes: whether companies can use books and other copyrighted works to build commercial models without paying the people who created them. Here, the authors’ allegation was specific. Anthropic used pirated copies of their books as training material for Claude, and the settlement puts a substantial financial value on that alleged use.
The amount is striking not only because of its size, but because it is tied directly to individual books. A payment of roughly $3,000 per book creates a much more concrete accounting than the usual arguments about datasets, tokens, or abstract model capabilities. For the authors covered by the agreement, the settlement connects the value of the training material to the works that supplied it.
It also places Anthropic’s use of pirated copies at the center of the resolution. The settlement does not describe copyrighted books as incidental material somewhere in a vast technical pipeline; it addresses them as identifiable works whose use led to payments for their authors. That distinction is likely to matter as other disputes over AI training move through the courts and companies have to explain not only what their models learned from, but how they obtained the material in the first place.
Euronews describes the $1.3 billion settlement in euro terms as the biggest copyright settlement on record, and likewise reports that it covers authors whose pirated books were used to train Claude. The dollar and euro figures reflect the same settlement being reported in different currencies; the underlying event is the judge’s approval of Anthropic’s agreement with the authors.
That makes this more than a large legal bill for one AI company. It is a landmark payout for the proposition that books used to train a commercial model can carry a measurable cost when the material was obtained through piracy. Claude may have been built to turn written language into a general-purpose system, but the settlement is a reminder that the inputs still came from somewhere — and that the people who created them may ultimately be part of the bill.