A U.S. federal judge in San Francisco approved a $1.5 billion [1] settlement on Monday between AI company Anthropic and a class of authors.
The decision ends a major legal battle over how artificial intelligence models are trained. It establishes a massive financial precedent for creators whose copyrighted works are used to build large language models without consent.
The class-action lawsuit accused Anthropic of illegally downloading and using books to train its Claude AI system [2]. By agreeing to the settlement, the company avoids further litigation regarding the origins of its training data [3].
An unnamed U.S. federal judge said the deal is the largest copyright class-action settlement in history [4]. The ruling provides finality to the authors' claims while allowing Anthropic to move forward with its operations.
While the primary settlement was approved, the judge did not grant the full amount requested for legal representation. The judge said the requested attorneys' fee award was reduced by $86 million [5]. This cut follows an original request for $187 million [5] in legal fees.
The settlement resolves the specific conflict between these authors and the developer of Claude, though it does not set a universal rule for all AI training practices. The court's approval on July 20, 2026 [6], marks the conclusion of this specific legal challenge in the U.S. District Court [7].
“The deal is the largest copyright class-action settlement in history.”
This settlement signals a shift in the AI industry from a 'move fast and break things' approach to a model where developers may need to pay for high-quality training data. By settling for $1.5 billion, Anthropic avoids a potential court ruling that could have legally redefined 'fair use' in the context of AI training, which would have created broader risks for other AI labs.
