AI

Anthropic Settlement: Judge Approves $1.5B Copyright Deal

A federal judge has granted final approval to Anthropic's $1.5 billion settlement with authors and publishers over pirated books used to train Claude.

By ExstarHub Team
Federal courthouse exterior illustrating final approval of the Anthropic copyright settlement.

A federal judge has granted final approval to the Anthropic settlement resolving claims over pirated books used to train the Claude chatbot. The $1.5 billion agreement is expected to pay authors and publishers about $3,000 for each eligible claimed book.

U.S. District Judge Araceli Martinez-Olguin approved the class-action settlement after describing it as meaningful relief for affected copyright holders. About 91% of the more than 482,000 books covered by the agreement have been claimed.

What the $1.5 billion settlement covers

The lawsuit focused on books Anthropic acquired through pirate websites and added to a central digital library. The settlement compensates eligible copyright owners without reversing an earlier ruling that addressed a separate question: whether using lawfully obtained books to train an AI model can qualify as fair use.

Plaintiff attorney Justin Nelson called the agreement the largest known copyright recovery in history. The approximately $3,000 payment is calculated per eligible work, while final distributions depend on the approved claims process.

Training and acquisition were treated differently

The case produced an important distinction before the parties settled. Former U.S. District Judge William Alsup ruled that training AI models on copyrighted books was fair use in the circumstances before the court. However, he found that downloading millions of books from pirate sites to build a permanent library was not excused simply because some of those books were later used for training.

That means the final settlement should not be described as a blanket ruling that all AI training on copyrighted material is legal or illegal. The court separated the transformative use involved in model training from the way Anthropic obtained and stored the source material.

Who is covered by the agreement?

The case began in 2024 with authors Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson. It later became a class action covering authors and publishers whose eligible books appeared in the dataset at issue.

More than 482,000 books are covered, and roughly 91% had been claimed by the time of final approval. The high claim rate clears the way for most of the settlement fund to reach identified rights holders after fees, expenses, and the court-approved distribution process.

Why the Anthropic settlement matters

The agreement puts a substantial financial cost on obtaining copyrighted training material through unauthorized libraries. AI developers may still argue that model training is transformative, but that argument does not automatically protect how they acquire or retain copyrighted works.

The case also gives authors and publishers a concrete recovery rather than leaving the dispute to years of additional litigation. Because this is a settlement, it does not create a final appellate rule for every future AI copyright case, but it will influence how AI companies assess dataset provenance and licensing risk.

What happens next?

Settlement administrators can move toward distributing payments to approved claimants. Anthropic resolves the class claims covered by the agreement, while the earlier fair-use analysis remains an important but case-specific part of the broader legal debate over generative AI training.

Key takeaways

  • A federal judge granted final approval to Anthropic’s $1.5 billion copyright settlement.
  • More than 482,000 books are covered, and about 91% have been claimed.
  • Eligible authors and publishers are expected to receive about $3,000 per book.
  • The case distinguished AI training from acquiring books through pirate websites.
  • The settlement resolves these class claims but does not settle every legal question about AI training data.

FAQ

Did the court rule that AI training on books is illegal?

No. An earlier ruling found the training use at issue to be fair use while treating Anthropic’s acquisition and storage of pirated books separately.

How much will authors receive?

The agreement provides about $3,000 for each eligible claimed book, subject to the final distribution process.

Source: AP News

Leave a Reply

Your email address will not be published. Required fields are marked *