Houthis Used Claude AI in Attempt to Develop Ballistic Missile Technology, Anthropic Says

An AI company has revealed that a weapons-development cell in northern Yemen, believed to be linked to the Iran-backed Houthi movement, used Anthropic’s Claude AI to help develop software for guided rockets and advanced missile systems.

According to Anthropic’s latest threat intelligence report, the group was involved in three weapons programmes, including a multi-stage ballistic missile with a stated range of more than 2,000 kilometres and a missile programme that included a hypersonic-glide variant.

Anthropic said the group used Claude Code to develop guidance, navigation and control software — the systems responsible for helping a flying weapon steer, stabilise itself and follow its intended trajectory.

The actors reportedly used several Claude sessions simultaneously, effectively assigning different AI instances different engineering roles. One was used for coding, another for research and another for reviewing the resulting software.

The group also used Claude to integrate an open-source autopilot with a phone-class flight computer and to conduct simulations related to missile flight and control.

Despite the extensive use of AI, Anthropic said it does not have evidence that the group successfully fielded an operational missile system. However, the company said the actors conducted a live test of a guided rocket in Yemen, which appears to have failed. Within hours of the failed test, the group reportedly returned to Claude to analyse what had gone wrong.

Anthropic said it subsequently identified and banned accounts connected to the operation and shared information with relevant public- and private-sector partners. The disclosure is part of a broader Anthropic report examining attempts by different actors to misuse Claude for military and other harmful activities.

The company said the Yemen case demonstrates how advanced AI can potentially lower the technical barriers involved in developing sophisticated weapons, particularly when users already possess relevant hardware and technical knowledge.

Anthropic also said the actors attempted to bypass its safeguards by disguising their objectives, splitting their work across multiple sessions and using other techniques to prevent individual conversations from revealing the full purpose of the project.

The revelations have added to growing concerns over the use of commercially available AI systems by militant groups and state-linked actors for military applications.

While the reported Yemen programme did not result in a confirmed operational ballistic missile, the incident shows that AI tools are increasingly being tested as substitutes for parts of traditional engineering and weapons-development workflows.