Anthropic has agreed to a proposed $1.5 billion settlement over claims it trained its Claude AI on pirated books. The agreement, pending court approval, is among the largest copyright settlements in U.S. history. It covers about 500,000 works at an average of $3,000 per work,
and requires Anthropic to destroy the disputed datasets.
In return, Anthropic will receive a release of liability for actions up to August 25, 2025. However, the deal does not shield it from future
claims involving reproduction, derivative works, or Claude’s outputs. The release applies only to the specific list of works agreed to by the parties