Yemeni forces used Claude to help develop a missile, Anthropic says
Some military personnel in northern Yemen, reportedly linked to the Houthis, used the AI model instead of engineers: Claude wrote code for the missile guidance system, configured the autopilot, simulated the flight and assembled the firmware.
They bypassed Anthropic’s safeguards by hiding the connection to weapons and breaking the development process into seemingly harmless tasks.
One of the missiles was even tested but crashed. A few hours later, the military personnel returned to Claude to figure out what had gone wrong.