Anthropic blocked Claude accounts used by unidentified actors in Houthi-controlled northern Yemen after detecting attempts to develop advanced missile systems, the company said in a report released Thursday. Anthropic said the users conducted a failed guided-rocket test but did not succeed in “fielding an operational device.”
According to Anthropic, the users returned to the Claude chatbot to investigate why the rocket test failed. The company said it identified a cell pursuing three weapons programs, including a multi-variant missile designed to glide at hypersonic speed and a warhead using mobile-phone hardware to maneuver during flight.
Anthropic said the actors used Claude Code in place of human software engineers to develop guidance, navigation and control software. Before the accounts were banned, the users had also built an offline simulation toolkit that did not depend on Claude or other computing platforms, according to the company.
The report did not identify the users. Northern Yemen is controlled by the Iran-backed Houthis, who already possess drones, missiles and other munitions. Hazam al-Assad, a member of the Houthis’ political bureau, rejected the suggestion that the group would depend on publicly available tools to build military capabilities and said its weapons were used for self-defense.
Trevor Ball, a weapons analyst at Armament Research Services, said the Houthis could be researching hypersonic missiles through Claude but lacked the production and technical capacity to build them. He noted that U.S. hypersonic missiles remain in testing and said the Houthis already possess Iranian-made anti-ship missiles capable of adjusting their guidance during flight.
The disclosure was Anthropic’s third report since March 2025 on global misuse of its AI platforms. Covering activity from December through August, it also described state-sponsored groups using the technology to spread propaganda and unnamed actors researching ways to make biological weapons more deadly.