Houthi Rebels Used AI to Develop Guided Missiles, Anthropic Admits

Anthropic, a leading artificial intelligence company, has disclosed that Houthi rebels have used its Claude AI to develop advanced weaponry, including self-correcting rockets and ballistic missiles. The Yemen-based group reportedly employed the AI in place of human software engineers to create guidance, navigation, and control systems for their weapons.

According to an Anthropic report, the threat actors accessed vast amounts of technical knowledge, including information from leading rocket scientists, through Claude. They used the AI to tune control settings, estimate position, and adapt missile firmware. The report noted that safeguards were in place to block many requests, but not all, and that the actors used tactics to circumvent security measures.

While Anthropic does not have evidence that the actors successfully deployed an operational device, they did conduct a test firing of a guided rocket, which appeared to fail. The rebels then used Claude to analyze the failure and make improvements. Evidence suggests they also developed an offline simulation toolkit independent of AI platforms.

The admission comes as the Houthis have escalated attacks in Saudi Arabia. Recent strikes on King Khalid International Airport resulted in fatalities and forced the airport's closure, prompting discussions in Washington about increasing U.S. military support for the Kingdom. The Houthis have also targeted shipping lanes, contributing to rising international oil prices.

Saudi Arabia's energy minister has vowed a strong response to the attacks on its airports and oil facilities. Security officials in Saudi Arabia have expressed shock at the range and accuracy of the Houthi warheads, which they attribute, in part, to the rebels' newfound AI capabilities. The development highlights the growing concern over the potential misuse of artificial intelligence in geopolitical conflicts and the redrawing of the global security landscape.