A US federal judge has approved Anthropic's $1.5 billion settlement with a group of authors who accused the AI company of using pirated copies of their books to train its Claude chatbot. The settlement, first agreed last year but initially rejected by the court, marks the largest payout yet in the wave of copyright lawsuits against AI firms and sets a significant precedent for how publishers and writers may seek compensation from tech companies training large language models on their work.

The lawsuit, filed by authors in 2024, centred on Anthropic's use of works from more than 480,000 books. The original judge, William Alsup, ruled that training Claude on the material counted as fair use, but found Anthropic liable for storing some 7 million pirated books in a central library, and initially rejected the settlement over concerns it was being forced on authors without enough detail. US District Judge Araceli Martinez-Olguin, who took over the case, has now approved the revised terms, under which authors will receive roughly $3,000 per affected work and Anthropic must destroy the pirated copies; some plaintiffs deemed the payout insufficient and opted out to pursue separate legal action.