Anthropic says Yemen’s fighters used AI to seek guided weapons

Yemeni forces reportedly attempted to employ Anthropic's Claude chatbot to create guided rockets and missiles, according to a report released by the US firm. The company detailed how it had identified and thwarted a cell of threat actors operating in northern Yemen, who were engaged in three weapons development programs. These programs focused on developing a guided rocket with final-phase homing guidance, a multi-stage ballistic missile capable of reaching over 2,000km (1,250 miles), and a multi-variant missile incorporating a hypersonic glide vehicle variant.

The cell utilized Claude to develop guidance, navigation, and control software for these airborne weapons, which would have otherwise necessitated a team of engineers. While Anthropic stated they do not have evidence the actors successfully fielded operational devices, they did conduct a test fire of a guided rocket, which allegedly failed shortly after.

The company promptly banned the involved accounts following the detection of the illicit activity. The Houthis, who hold control over much of northern Yemen, including Sanaa, are believed to be the entities behind these attempts. The report comes amidst a series of offensive actions by the Iran-backed militants across Yemen, with recent claims of control over a Red Sea island near the Bab al-Mandab strait, a crucial shipping lane connecting Europe to Asia.