Anthropic has disclosed a serious security incident in which a Houthi-linked weapons cell used its AI model Claude to develop three advanced weapons programs simultaneously – the first documented case of generative AI being leveraged for strategic weapons development.
According to the investigation, the cell used Claude Code to build guidance, navigation, and control systems across three programs:
· A guided rocket.
· A ballistic missile with a range exceeding 2,000 kilometers.
· A hypersonic glide vehicle.
Records show the cell conducted a test-firing of the guided rocket using AI-generated guidance software. When the test failed, they returned to Claude within hours to diagnose the failure and improve performance.
Anthropic confirmed it identified and banned the accounts, stressing that no operational weapon was successfully deployed and that its security systems successfully detected the misuse.
Security experts warn this incident highlights the dual-use risks of AI, where civilian technologies can be repurposed for lethal weapons development, posing new challenges for tech companies and regulators.