Court approves one copyright case, but AI training fight continues

A federal judge in San Francisco approved Anthropic’s $1.5 billion settlement with authors who said the company trained its Claude chatbot on pirated books, closing one of the biggest copyright cases to hit the AI industry.

The deal ends Bartz v. Anthropic PBC, the class action filed Aug. 19, 2024 by authors Andrea Bartz, Charles Graeber and Kirk Wallace Johnson. Anthropic’s settlement covers authors and publishers whose books were included in the case, and company counsel said more than 91% of the authors and publishers covered by the agreement had claimed their share.

What the settlement resolves is Anthropic’s financial exposure in this case. What it does not resolve is the central legal question that has shadowed every major AI developer: whether copyrighted books can be used to train models without permission, payment or a license. The court had already issued a summary-judgment ruling finding that training large language models on copyrighted materials was fair use, but the settlement stops the dispute short of a final appellate ruling that could have carried far beyond Anthropic.

That missing precedent is the key reason the settlement matters so much to publishers, authors and other creators. A $1.5 billion payout sets a powerful price tag on alleged misuse of books, yet it leaves no binding answer for OpenAI, Meta, Google or other companies building on large text corpora. The industry still faces the same unresolved mix of copyright risk, licensing pressure and litigation exposure, even as one of the highest-profile cases reaches the finish line.