Judge approves $1.5B Anthropic settlement over pirated books

In short: A federal judge approved a $1.5 billion settlement in which Anthropic will pay authors and publishers for pirated books used to train its Claude chatbot.

What happened

A federal judge approved a $1.5 billion settlement to resolve a class-action lawsuit against Anthropic. The lawsuit said the company used pirated copies of books to help train Claude, its chatbot.

The case focused on books allegedly taken from Library Genesis and Pirate Library Mirror, which are known for sharing unauthorized copies. The approved class covers about 500,000 book titles after removing duplicates and other works that did not qualify.

Under the settlement, payments work out to about $3,000 per book. The Associated Press reported that around 91% of the more than 482,000 books covered by the ruling had already been claimed by eligible authors or publishers.

The court said the deal provides “meaningful relief” for people who own the rights to the books. The judge also separated two issues, training on lawfully obtained books, which he had earlier suggested could be allowed under fair use, and piracy, which the court treated as a separate copyright problem.

Why it matters

This deal shows that courts may treat “how the book was obtained” as a key question, not just whether a company used books to train an AI system (like using many books as practice examples, the way a student learns by reading). It also matters for authors because the $3,000 figure is not guaranteed for everyone. If a book has multiple rightsholders, such as an author and a publisher, the payment is shared, so some people may receive less.

Source: NYTimes