What Anthropic's $1.5bn book settlement means for you

A US judge has signed off on Anthropic's $1.5 billion settlement with authors and publishers over books used to train its Claude models, in what's believed to be the largest copyright settlement in US history.

On Monday 20 July 2026, US District Judge Araceli Martínez-Olguín granted final approval to the settlement, which resolves a class action brought by authors whose books were used to train Claude, Anthropic's AI assistant. Payments to authors are now beginning, working out to about $3,000 per work across roughly 500,000 books, according to TechCrunch.

The dispute centred on how Anthropic got hold of the books in the first place. The company obtained training material two ways: buying and scanning physical copies, and downloading books from piracy sites. A court had already ruled that training AI on copyrighted text counts as fair use, a legal doctrine that allows limited use of copyrighted material without permission. But it also ruled that downloading books from piracy sites was not covered by that protection. Rather than go to trial over the piracy claims, Anthropic settled.

The settlement closes this particular case, but it doesn't set a binding legal precedent for the wider AI industry. Other copyright lawsuits over training data remain pending against Google, Meta, Midjourney and OpenAI. In July, publisher Hachette and author Scott Turow also sued Google over the data used to train its Gemini models.

If you're an author whose book was part of the roughly 500,000 works covered, you're in line for a payout of around $3,000 as part of this settlement.

For everyday users of Claude or other AI tools, the settlement doesn't change how the products work today. But it does put a concrete price on pirated training data, which may push Anthropic and other AI labs towards buying or licensing books rather than scraping them from piracy sites in future.

The other pending cases against Google, Meta, Midjourney and OpenAI will test similar questions about how AI companies source their training data. Because this settlement sets no binding precedent, those cases could still produce different outcomes, in and out of court.