Yemen group sought guided weapons with Claude, Anthropic says

A group in northern Yemen tried to use Anthropic's Claude chatbot to build guided rockets and missiles, the US artificial intelligence company said in a report on illegal uses of its systems.

Anthropic said it had “identified a cell of threat actors based in northern Yemen running three weapons development programs”. It said the programs involved a guided rocket using a phone-class flight computer with final-phase homing guidance, a multi-stage ballistic missile with a stated range goal above 2,000 kilometres, and a missile with a hypersonic glide vehicle variant.

The company said the cell used Claude to develop guidance, navigation and control software that would otherwise have required a team of engineers. Anthropic said it had no evidence the actors succeeded in fielding an operational device, but said they test-fired a guided rocket that appeared to fail. The company said the actors returned to Claude within hours to work out why it failed.

Anthropic said it banned the relevant accounts as soon as it detected the activity. The company did not identify the actors.

The Yemen case appeared in the same Anthropic report that described other alleged misuse of Claude by state-sponsored or state-aligned actors from China, Iran and west Africa in surveillance operations. The company said it found and disrupted incidents between January and July.

Anthropic also alleged that Chinese developers deceptively used Claude to answer user queries while using that data to improve their own models, and said it disrupted other efforts involving weapons design, questionable biological research and dating scams.

The Yemen disclosure comes as Iran-backed Houthi rebels are on the offensive in the country. The Houthis control much of northern Yemen, including the capital, Sanaa, and on Friday claimed control of a Red Sea island in the Bab al-Mandab shipping lane.