Anthropic has disclosed that Iran-linked actors and Houthi rebels in Yemen used its Claude AI model for weapons development and military surveillance, including work against U.S. naval forces in the Middle East. The company said it banned the accounts involved, added new detection mechanisms, and passed its findings to government authorities. The disclosures cover three distinct sets of activity tracked by the AI developer.
In one case, an Iran-linked threat actor used Claude to support military reconnaissance and to develop targeting recommendations against U.S. naval forces. The operation combined publicly available ship and aircraft transponder identifiers with commercial satellite imagery and data on U.S. naval movements, and it also pulled the names of U.S. military personnel from captions on publicly available military photographs. The actor additionally researched possible vulnerabilities in communications equipment aboard ships, including known flaws in Cobham Sailor VSAT terminals, Cisco communications equipment, and Schneider Electric EcoStruxure systems.
A second case involved a Houthi-controlled cell in northern Yemen that used Claude Code across three weapons programs. Those were a guided rocket using a phone-class flight computer to assist terminal guidance, a multistage ballistic missile with a range of more than 2,000 km, and the R2000 missile family, which included a hypersonic glide vehicle variant. According to Anthropic, the group used Claude to write guidance, navigation, and control software; integrate an open-source autopilot with a phone-class flight computer; write control and position-estimation code; tune parameters; build firmware; and run flight simulations. The cell effectively substituted multiple Claude instances for a team of software engineers handling coding, code review, research, and simulation.
The Houthis are not Iranian, but the disclosure notes they can share research and development results with their allies and potentially draw on Iran's industrial capacity to build weapons. Iran also used Claude for surveillance tools beyond military targeting. One Iran security-linked unit used the model to analyze 155,216 tweets to profile, identify, and surveil 6,388 opposition individuals in a single year. Another group used it as an engineering pipeline to develop domestic tracking tools, including the production-deployed "al-Najm al-th qib" Firefox extension, built to mass-harvest user identities across major social platforms.
Anthropic has banned 16 Claude accounts tied to Iranian paramilitary and domestic security agencies, though that does not mean all such accounts have been banned. The company also said Iran-linked actors and the Houthis are not alone in putting Claude to military use, stating that China and Russia are actively using the model for their military programs as well. The findings raise questions for U.S. policymakers and AI developers about how American models can be repurposed by foreign military and security organizations.
More AI news from TechManNews.
