In 2025, a federal judge handed down one of the most consequential copyright rulings in AI history. Anthropic (Claude AI) had purchased and physically destroyed over 2 million books as part of a strategy to acquire AI training data. The judge ruled it legal.