3 Sources
[1]
Users in Houthi-held Yemen tried to develop advanced weapons with AI, Anthropic says
BEIRUT (AP) -- Anthropic says Claude users in northern Yemen, territory controlled by Iran-backed Houthi rebels, tried to use the AI model to develop advanced missiles. AI is already transforming warfare from Ukraine to Gaza, and its use on a rugged and remote battlefield is likely to increase
[2]
Anthropic says Yemen fighters used AI to seek guided weapons
An AI firm revealed a Yemen group used its chatbot for weapons development. They attempted to build guided rockets and ballistic missiles with advanced capabilities. The group tested a guided rocket which appears to have failed during its flight. Anthropic banned the accounts after detecting this
[3]
Iran-backed rebels in Yemen tried using Anthropic's Claude to build guided missiles, alarming report finds
Iran-backed Houthi rebels in Yemen tried to use Anthropic's powerful Claude AI model to craft software for use in ballistic missiles, according to an alarming report released by the company. Anthropic did not identify the Houthis by name - instead referencing "a cell of threat actors based in
Share
Copy Link
Anthropic revealed that users in Houthi-held Yemen attempted to use its Claude AI model to develop advanced weapons including guided rockets and ballistic missiles. The company blocked the accounts after detecting the illegal activity, though a test-fire of a guided rocket had already failed.

Anthropic disclosed Thursday that users in northern Yemen attempted to exploit its Claude AI model to develop advanced weapons, marking a concerning escalation in AI-enabled weapons development across conflict zones
1
. The company identified a cell of threat actors running three distinct weapons programs in the region controlled by Iran-backed rebels, though it did not explicitly name the Houthis2
. The revelation comes as AI continues transforming warfare from Ukraine to Gaza, raising urgent questions about the technology's proliferation in remote battlefields.The actors pursued development of a guided rocket using commodity phone-class flight computers with final-phase homing guidance, a multi-stage ballistic missile with a stated range goal above 2,000 kilometers, and a multi-variant missile that included a hypersonic missiles variant
2
. Anthropic stated the users did not succeed in fielding an operational device but did conduct a test-fire of a guided rocket that appears to have failed1
. Within hours of the failed test, the actors returned to Claude to troubleshoot why their guided weapons malfunctioned, providing Anthropic with evidence of the attempted misuse.The cell used Claude Code instead of human software engineers to develop guidance, navigation and control software for the airborne weapons
2
. To circumvent Anthropic's built-in safeguards, the malicious actors tried hiding their goals and the products the software was meant for, while breaking up requests across multiple sessions3
. Anthropic acknowledged its safeguards blocked many requests but not all of them. The company banned the relevant accounts as soon as it detected the illegal activity and shared its findings with private and public partners1
.Before the accounts were banned, Anthropic found evidence that the users had already built an offline simulation toolkit that doesn't use Claude or any other computing platforms
1
. This suggests the actors may continue development efforts independently. Trevor Ball, a weapons analyst at Armament Research Services, noted that while the group might be exploring hypersonic technology by asking Claude, they lack the production or technical capabilities to actually build such weapons, as even U.S. hypersonic missiles remain in testing1
. Ball suggested the actors are likely trying to develop their own capabilities to reduce reliance on Iranian shipments of weapons and components.Related Stories
The Yemen case formed part of Anthropic's third report on global AI misuse since March 2025, covering findings from December to August
1
. The report detailed several other alarming episodes including an Iran-linked group requesting information on US military targets, including records on American ship and aircraft movements and weaknesses in communications systems3
. Anthropic also disrupted efforts by scientists to use its models for biological weapons research and identified Russia and China-based individuals drafting plans for conventional weapons3
. Additionally, the company uncovered Russia-linked actors using Claude for spreading propaganda and efforts in China and Iran to spy on dissident groups.The revelation coincides with Iran-backed militants launching a wide-ranging offensive across Yemen toward the strategic Bab al-Mandab strait, which connects Europe to Asia
2
. The Houthis claimed control of a Red Sea island in the middle of the Bab al-Mandeb shipping lane, completing their push to control the area. The rebels have made lightning territorial gains since Wednesday in their war with Yemen's Saudi-backed government, seizing strategic territories along the Red Sea coast1
. Adam Baron, a Yemen-focused researcher at the New America think tank, emphasized the group's technological sophistication, stating there's a tendency to see the Houthis as barefoot tribal fighters when they actually demonstrate incredible institutional tech savvy through their use of Claude, social media manipulation, and ability to capitalize on Iranian expertise transfer1
. The incident underscores growing concerns about weaponized AI and the risks ranging from human misuse to AI models themselves causing harm, with former Anthropic and OpenAI researcher Jacob Coxon warning that companies are racing straight to self-improving superintelligence and gambling with our lives3
.Summarized by
Navi
28 Feb 2026•Policy and Regulation

08 Sept 2026•Policy and Regulation

11 Mar 2026•Technology

1
Science and Research

2
Policy and Regulation

3
Technology