Houthi rebels responsible for deadly missile strikes in Saudi Arabia have been using Anthropic’s artificial intelligence to make warheads, the US firm has admitted.
The makers of Claude revealed in a report last month that the Yemen-based fighters launched three weapons programmes using their AI products.
Over the weekend, the Houthis launched their most devastating attacks on Saudi Arabia to date.
The multi-billion pound company’s report provides the clearest evidence yet of how AI could redraw the geopolitical landscape in the Gulf and elsewhere.
In a section entitled ‘Disrupting a Yemen-based guided weapons engineering cell’ the authors described how they identified a terrorist cell running three programmes, all using Claude.
Often fighting barefooted or in sandals, the Houthi appear a primitive force. However, thanks to Claude the Iranian-backed guerillas have learned to construct long range rockets so clever they correct their flight if they are heading off target.
It is unclear whether the Houthi used Claude as part of their attacks on King Khalid Airport but the strikes have stunned the Saudis and triggered discussions in the White House about the US increasing its military support for the Kingdom.
Houthi gunmen are also intent upon disturbing shipping transiting through the Suez Canal in a bid to damage Saudi Arabia’s economy. Their efforts are pushing up international oil prices as the Strait of Hormuz has been in effective lockdown for months due to the US-Iran conflict.
Claude is capable of a wide variety of conversational and text-based teaching programmes on almost all areas of human activity.
In this case the AI engine condensed the knowledge of the world’s leading rocket scientists, which was then accessed by terrorists, the Anthropic report said.
It added: ‘Threat actors based in northern Yemen [were] running three weapons development programmes: a guided rocket that used a commodity, phone-class flight computer with final phase homing guidance, a multi-stage ballistic missile with a stated range goal of more than 1,500 miles and a multi-variant missile, referred to as the R2000, that included a hypersonic glide vehicle variant.’
The threat actors, understood to be Houthis, used Claude Code in place of human software engineers to develop the guidance, navigation and control software that steers and stabilises a flying vehicle.
Claude Code provided position estimation software, tuned the control settings, ran a firmware build pipeline and when warheads missed their targets adapted itself to improve their accuracy.
Even when Anthropic realised their software was being used in this manner the threat actors were able to take what they had been taught and make adaptations.
The Anthropic report continued: ‘Our safeguards blocked many of their requests, but not all of them. They used a variety of tactics to evade our safeguards and split their work across multiple sessions.
‘We do not have evidence the actors succeeded in fielding an operational device; but they did test-fire a guided rocket. This field test appears to have failed. Within hours, the actors returned to Claude to work out why it failed.
‘We have evidence that the actors had already built an offline simulation toolkit that does not rely on Claude or other engineering computing environments such as MATLAB.’
Saudi Arabia’s energy minister has vowed a ‘serious’ response to the deadly attacks on the Kingdom’s airports in recent days and oil and gas facilities.
Saudi security officials have been shocked by the range and accuracy of the Houthis’ warheads.
There have been two attacks on the King Khalid International Airport, killing 14 people. The attacks forced the closure of the airport.
