Houthis used Anthropic AI to develop ballistic missile software, says report

A Yemen-based threat actor cell attempted to leverage Anthropic's Claude AI model to develop guidance, navigation, and control software for ballistic missiles, according to a recent US artificial intelligence company intelligence report. The Financial Times reported that the cell, operating from northern Yemen under the control of the Houthi movement, was working on three weapons development programs: a multi-stage ballistic missile, a multi-variant missile, and a guided rocket. The cell aimed to utilize a phone-class flight computer in their endeavors.

Anthropic stated that its safeguards managed to block "many of their requests," but not all, as the threat actors attempted to bypass these defenses by concealing their intentions and fragmenting their work across multiple sessions. This report marks a detailed intelligence assessment of various attempts to misuse Claude for cyberattacks, influence operations, surveillance, biology, and weapon development. Anthropic's team successfully disrupted all operations detailed in the report.