< Back to all clusters
[TECHNOLOGY] · Yemen, Iran, United States · 5 sources

started · updated

Anthropic reports Houthi-linked actors used Claude AI for missile development

Anthropic has reported that actors in northern Yemen, linked to Houthi militias and potentially Iranian interests, attempted to use its Claude AI model to develop advanced weaponry. According to the company, these users sought to create missile guidance and flight control software, including projects involving guided rockets and ballistic missiles.

To bypass safety guardrails, the actors reportedly used a technique of decomposing complex requests into multiple, seemingly innocuous sessions. This method prevented the AI from detecting the overarching malicious intent. While Anthropic stated there is no evidence that combat-ready weapons were successfully produced, they noted that one failed test flight may have occurred.

Reports also suggest that these actors used the AI to gather intelligence and improve targeting capabilities against US Navy vessels by analyzing public data and communication system vulnerabilities. In response, Anthropic has blocked the implicated accounts and shared threat intelligence with partners. Ansar Allah officials have dismissed the allegations, claiming their military capabilities are developed through their own advanced technologies rather than public AI resources.

Entities

Anthropic · Claude · Houthi · Houthi militia · Iran · US Navy · Yemen

Claims

What the coverage asserts, and how many sources carry each claim.