Anthropic says users operating from Houthi-controlled northern Yemen attempted to use its Claude system for work involving hypersonic missiles and guided warheads. The company says its safeguards prevented the users from completing their objectives and their accounts were blocked.
The report moves the AI-safety conversation beyond misinformation and workplace disruption into potential weapons development.
What is confirmed
This is Anthropic’s account of activity detected on its platform. The company has not publicly identified the users, and there is no evidence that Claude produced a deployable weapon.