Key Highlights
- Anthropic identified a threat actor cell based in northern Yemen, an area largely controlled by Houthi rebels, running three separate weapons development programs using its Claude AI model.
- The group used Claude Code in place of human software engineers to develop guidance, navigation and control software for a multi-stage ballistic missile, a multi-variant missile, and a guided rocket using a Commodity phone-based flight computer.
- Anthropic said its safeguards blocked many but not all of the group's requests, with the actors evading detection by disguising their goals and splitting their work across multiple sessions.
- A field test of the guided rocket reportedly failed, though Anthropic said it has evidence the group had already built an offline simulation toolkit that does not depend on Claude or other AI-assisted engineering tools.
A Yemen-Based Cell Using AI in Place of Engineers
Yemeni militants sought to use Anthropic's Claude AI model to develop guidance, control, and navigation software for ballistic missiles, according to the company's latest intelligence threat report. Anthropic did not directly name the Iran-backed Houthi movement in its disclosure, but said it identified a cell of threat actors based in northern Yemen, an area largely under Houthi control, running three separate weapons development programs.
The report highlights the growing risk of advanced AI capabilities falling into the hands of unauthorized actors as the underlying technology continues to improve rapidly. The same Anthropic report separately disclosed multiple attempts by researchers to use the company's AI model for biological weapons research. AI labs currently operating at the frontier of the technology face an ongoing struggle to prevent unauthorized use of their systems in ways that could threaten the United States and its allies.
How the Group Used AI to Replace Human Engineers
Anthropic said the Yemeni cell sought to develop several distinct weapons, including a multi-stage ballistic missile, a separate multi-variant missile, and a guided rocket built around what the company described as a commodity phone-class flight computer. According to the company's report, the actors used its Claude Code product in place of human software engineers to develop the guidance, navigation, and control software needed to steer and stabilize a flying vehicle.
The group reportedly used the AI model to integrate an open-source autopilot system onto the phone-based flight computer, writing the associated control and position estimation software, tuning control settings, running a firmware build pipeline, and performing flight simulations, effectively substituting AI assistance for specialized engineering expertise the group may otherwise have lacked.
Safeguards Blocked Many but Not All Requests
Anthropic said its safeguards blocked many of the group's requests but not all of them, with the actors employing a range of tactics specifically designed to evade the company's detection measures. According to the report, the group disguised the underlying goals of their work and the intended products the resulting software was meant to power, and split their efforts across multiple separate sessions so that no single interaction fully revealed the complete scope of their intent.
The disclosure comes as the Houthi movement, designated a terrorist group by the United States and considered one of the most capable members of Iran's broader network of allied militant groups, launched a renewed offensive along Yemen's Red Sea coast this week, capturing a strategic port city and advancing on the Bab al-Mandeb Strait, a critical waterway for international shipping. The group's advance comes alongside intensified attacks on Saudi Arabia's energy infrastructure and a reignition of Yemen's civil war, which has continued for more than a decade. The Houthi movement's advance toward the strait, which connects the Red Sea to the Gulf of Aden and the wider Indian Ocean, has been cited as a contributing Factor in pushing Brent Crude prices above USD 105 a barrel.
A Long-Running Military Buildup With Iranian Backing
Analysts say the Houthi movement, which has controlled Yemen's Capital and much of the country's populous northern region since 2014, has been steadily developing its missile capabilities over time, receiving weapons and technology support from Iran. Its advance toward the Bab al-Mandeb Strait represents a significant setback for Saudi Arabia, which intervened in Yemen's civil war in 2015 to lead a regional coalition against the Houthis, and has bolstered Iran's broader strategic position as it seeks to use higher energy prices as Leverage in its ongoing conflict with the United States.
The Houthi movement had largely remained outside the direct US-Israel war against Iran, with the exception of firing occasional missile and drone salvos toward Israel. A 2022 truce between the group, Saudi Arabia, and the Yemeni government had held until July, when the Houthis declared a blockade on Saudi ports and began attacking Saudi oil tankers. The Bab al-Mandeb route has become an increasingly vital corridor for Saudi oil exports since Iran began restricting shipping through the Strait of Hormuz following the outbreak of the broader war against Iran in February.
A Failed Test, but a Continuing Threat
Anthropic said the Yemen-based actors carried out a sustained effort to develop guided weapons, including using its AI model to design guidance software, though the company said it lacks evidence the group succeeded in fielding a fully operational device. The group did test-fire a guided rocket, an effort that Anthropic said appears to have failed, with the actors returning to the AI model within hours of the test to work out what had gone wrong.
Anthropic declined to confirm directly whether the actors it identified in northern Yemen were indeed the Houthi movement, though the company acknowledged few other groups plausibly fit that description. The company said it banned the accounts associated with the actors and shared relevant threat information with public- and private-sector partners to help mitigate the broader risk. Anthropic added that it has evidence the group had already built an offline simulation toolkit that does not rely on Claude or any other AI-assisted engineering environment, suggesting the underlying threat persists independent of the company's own safeguards.
Outlook
Whether continued restrictions on AI model access can meaningfully slow further weapons development by groups like the one identified in Yemen, or whether the existence of an offline simulation toolkit means such actors have already developed sufficient independent capability regardless of future safeguards, remains an open and consequential question as the broader regional conflict continues to escalate.






Please wait processing your request...