Houthis used Claude to build missiles that adjust after each strike, Anthropic admits

American artificial intelligence company Anthropic has revealed that Yemen-based militants used its Claude AI system to support three weapons development programmes, including guided rockets and ballistic missiles.

According to the Daily Mail, the suspected Houthi operatives exploited Claude Code to develop software for missile guidance, navigation and flight control.

Anthropic said the projects included a guided rocket, a ballistic missile with a proposed range exceeding 1,500 miles and a missile incorporating a hypersonic glide vehicle variant.

The company acknowledged that its security measures prevented several requests but failed to block others.

“Our safeguards blocked many of their requests, but not all of them,” Anthropic stated.

It disclosed that the militants tested a guided rocket unsuccessfully before returning to Claude for troubleshooting.

However, Anthropic emphasised that it had no evidence the group successfully deployed an operational weapon.

The revelations followed deadly attacks on Saudi Arabia’s King Khalid International Airport, although any connection between Claude and those strikes remains unconfirmed.