Anthropic accuses chinese open weight labs of theft, while it has had to pay $1.5B for theft.

Anthropic, the AI safety startup behind Claude, finds itself at the center of two major intellectual property controversies. The company has reached a staggering $1.5 billion settlement with a group of authors, including George R.R. Martin and John Grisham, who sued for copyright infringement related to the use of their books to train Anthropic's models. This settlement, one of the largest of its kind, sets a critical financial precedent for the cost of training data in the AI industry. Concurrently, Anthropic has levied accusations against several Chinese open-weight AI labs, alleging they illicitly obtained and replicated its proprietary model weights and architecture—the core 'recipe' of its AI systems—raising significant national security and competitive concerns.

The $1.5B settlement, detailed in an NPR report, resolves a class-action lawsuit and establishes a framework for future compensation, potentially forcing AI firms to budget billions for data licensing. This contrasts sharply with Anthropic's stance as an accuser in the alleged Chinese IP theft, where it claims rivals bypassed immense R&D costs by copying its technology. This duality underscores the industry's central hypocrisy: companies aggressively protect their own IP while building foundational products on potentially unlicensed copyrighted material. The outcome will influence global AI governance, data sourcing strategies, and the viability of the open-weight model movement.