Anthropic has won closure in one of the biggest copyright cases yet brought against an AI company, but the broader legal battle over how AI models are trained is far from settled. The new court-approved payout ends a dispute over allegedly pirated books while leaving the industry’s central fair-use argument intact.

A federal judge has now given final approval to Anthropic’s $1.5 billion class-action settlement with authors and publishers who said the company used copyrighted books without permission to build its AI systems.¹ The deal has been described as the largest copyright recovery on record, with payments of roughly $3,000 per covered work.²

For Anthropic and many participating claimants, the settlement offers a practical end to a costly fight. Reporting on the approval noted that more than 90 percent of affected authors and publishers had already filed claims, while only a small number opted out.³ The judge also trimmed attorneys’ fees, cutting the request down to about $101 million, a sign of the court’s effort to answer criticism that too much of the fund would go to lawyers rather than writers.⁴

But the settlement has never been free of controversy. Before final approval, the court temporarily put the deal on hold amid concerns that authors might be pressured into accepting terms negotiated behind closed doors and that notice to class members had to be clearer.⁵ Some authors and publishers also argued that $3,000 per book was too little, especially given the scale of the alleged copying.¹

The deeper tension is that this case did not decisively answer the question hanging over the AI industry. Earlier rulings suggested Anthropic’s use of lawfully obtained books for training could qualify as fair use, while its alleged downloading of books from pirate sources was treated as a separate and potentially illegal act.⁶ That means this settlement resolves one piracy dispute, not the wider question of whether training AI on copyrighted material is legal across the board.⁷