Houthis used Anthropic's Claude AI to help develop ballistic missiles and guided rockets, the artificial intelligence company disclosed in a September 2026 threat-intelligence report. Anthropic said a weapons-development cell operating in northern Yemen used its Claude Code coding tool in place of human software engineers to build the guidance, navigation and control software that steers and stabilises a flying vehicle. The disclosure, which resurfaced widely in international coverage on October 12-14, 2026, is the clearest documented case yet of a frontier AI system being woven into a real-world weapons programme.
How the Yemen Cell Used Claude Code
Anthropic did not name the Houthis directly, but identified a cell of threat actors based in northern Yemen — territory controlled by the Iranian-backed Houthi movement — running three weapons development programmes between December 2025 and August 2026.
The company described an engineering workflow that will look familiar to any software team. The cell ran multiple Claude Code instances in parallel — one writing code, another conducting research, a third reviewing the first instance's output. Anthropic said the actors concealed their intentions and split tasks across separate sessions, making it harder for any single interaction to reveal the full purpose. The AI helped integrate an open-source autopilot onto a commodity, phone-class flight computer: writing control and position-estimation software, tuning the settings, running a firmware build pipeline and performing a flight simulation.
Three Weapons Programmes, One Failed Test Launch
According to Anthropic, the three programmes were: a guided rocket built around a commercially available, phone-class flight computer with final-phase homing guidance; a multi-stage ballistic missile with a stated range goal above 2,000 kilometres; and a multi-variant missile family the actors called the R2000, which included a proposed hypersonic glide vehicle variant.
The group test-fired a guided rocket. The test appears to have failed — and within hours the actors returned to Claude to work out why, bringing post-test data in for diagnosis. Anthropic said it found no evidence that the cell fielded an operational weapon. It also noted that the group had compiled its simulation toolkit into a standalone executable able to run without Claude and without engineering environments such as MATLAB — so parts of the project could continue even after access was cut off.
Anthropic's Response and the Wider Misuse Report
Anthropic said it banned every account it could link to the cell and shared relevant threat intelligence with public- and private-sector partners. The company acknowledged that its safety controls blocked many of the actors' requests, but some prohibited activity still got through. The report also noted that scientists made multiple attempts to use Claude to research biological weapons.
The Yemen cell was one of six conventional-weapons operations detailed in the report. Parallel findings included Iran-linked actors using Claude to compile targeting recommendations involving US Navy warships — transponder data, commercial satellite imagery and naval-system vulnerabilities — alongside cyber operations and surveillance tooling. Responding to the Yemen findings, Houthi political bureau member Hazam al-Assad called it unreasonable and illogical to suggest the group would rely on open sources for military production, saying its forces have their own accumulated capabilities.
What the Disclosure Means for AI Safety
Anthropic found no evidence that any operational weapon used in real attacks was built with Claude's help — but the disclosure offers a concrete, hands-on window into how AI is changing the economics of weapons development for groups that lack large engineering staffs. The strategic lesson is scalability: accessible AI lowers the technical and time costs of weapons engineering while making platform-level safeguards and threat intelligence pivotal to keeping it out of the wrong hands.
Watch: Mint's report on how Houthis used Anthropic's Claude AI for weapons development — a video explainer walking through Anthropic's September 2026 misuse report, the three Yemen weapons programmes and how Claude Code replaced human software engineers.
Music in this post's reel: Kevin MacLeod - Darkening Developments (incompetech.com), licensed under CC BY 3.0
Sources
- The Statesman: "Houthis used Anthropic AI to develop ballistic missiles" (October 13, 2026)
- Asia News Network: "Yemen's Houthis used Anthropic AI to develop ballistic missiles" (October 14, 2026)
- Archyde: "Anthropic Disrupts Yemen Cell Using Claude Code for Weapons Software" — covering the standalone toolkit, Houthi response and the report's other misuse findings