Anthropic's $1.5B Copyright Settlement Gets Final Approval, But the Fair Use Question Lingers

Anthropic is now cleared to begin distributing $1.5 billion to authors and publishers whose copyrighted books were used without permission to train its AI models. A federal judge gave final approval to the class-action settlement on Monday, closing a case that had drawn intense scrutiny from across the tech and publishing worlds.

The settlement, believed to be the largest in U.S. copyright history, will pay roughly $3,000 per work across an estimated 500,000 titles. But the resolution leaves the industry's most consequential question unanswered: Is training AI on copyrighted content fair use?

Judge William Alsup, who oversaw the case before retiring, ruled that training an AI model on copyrighted text is fair use — a landmark decision that AI companies celebrated. However, he also found that Anthropic crossed a legal line by downloading books from pirate sites like Library Genesis and Pirate Library Mirror. That separate act of piracy, not the training itself, formed the basis of the liability that led to the settlement.

By settling, Anthropic ensured the fair use question will never be tested at the appellate level. Alsup’s ruling remains a single district-court decision with no binding authority on other judges. As a result, the flood of similar lawsuits against Google, Meta, Midjourney, and OpenAI — including a new class action filed last week against Google over Gemini training data — will continue to play out in different courts with potentially different outcomes.

For developers building on top of these models, the legal uncertainty is far from over. The safe harbor of fair use is not yet settled law, and the cost of training on unlicensed data could rise dramatically in the future. Anthropic’s settlement, while historic, may ultimately be remembered as the opening bid in a much longer negotiation.