AI firm Anthropic has agreed to a $1.5 billion settlement over claims it used copyrighted books to train its Claude model. The deal could compensate up to 500,000 authors $3,000 each if approved by a U.S. court.