Anthropic reaches $1.5 billion settlement with publishers over AI training data

A federal court has approved a $1.5 billion settlement between AI developer Anthropic and a class of authors and publishers who alleged that their copyrighted books were used without permission to train Anthropic’s Claude language model. About 97% of the class accepted the settlement, with only a few hundred authors opting out for separate litigation. Bloomsbury Publishing will receive a share of the payout, marking one of the largest copyright agreements involving AI to date.

The deal compensates past use of copyrighted material but does not include a forward‑looking licence, leaving future training data negotiations open. Industry observers view the settlement as a benchmark that assigns a price to unlicensed training data and signals a shift toward formal licensing arrangements for foundation models.