Anthropic: Iran, Russia, China, Yemen misused Claude AI for weapons development

Anthropic's threat intelligence report identifies state-linked users from Iran, Russia, China, and Yemen as having misused its AI models over the past year, including for weapons development. The report notes that the most advanced models (Fable and Mythos) were not involved, limiting the direct impact. This is a first public attribution by a major AI lab of state misuse for weapons, signaling potential regulatory and security implications.