Anthropic Says Yemen-Based Weapons Cell Used Claude in Missile

A weapons development cell based in northern Yemen used Anthropic’s Claude Code to support three weapons programs, including the development of guidance software, according to Anthropic’s September 2026 threat intelligence report, as cited by Coin Bureau.

Anthropic said the cell worked on a guided rocket program and two missile programs. The company identified the activity through its own threat investigations, subsequently banned the associated accounts and shared information with public- and private-sector partners. Anthropic said it found no evidence that the group successfully fielded an operational weapon.

Claude Code Used Across Multiple Weapons Programs

According to Anthropic, the Yemen-based actors used Claude Code to assist with guidance, navigation and control software for the weapons programs. The company said the users employed multiple Claude instances for different tasks, effectively using the AI system to support work that would ordinarily require software engineering expertise.

The programs included a guided rocket, a ballistic missile with a stated range goal above 2,000 kilometers and a missile program that included a hypersonic-glide variant. Anthropic said the actors attempted to use Claude across the programs simultaneously.

The company also reported that the group conducted a test of a guided rocket that appeared to fail. Within hours, the actors returned to Claude to investigate the failure, according to Anthropic’s account. Despite the activity, Anthropic said there financial was no evidence that an operational weapon was successfully deployed.

AI Safety Controls Face New Misuse Challenge

The case illustrates a growing challenge for AI developers: preventing capable coding and reasoning systems from being repurposed for dangerous applications while maintaining legitimate access for users.

Anthropic said its safeguards blocked many of the group’s requests, but not all. The actors attempted to evade those controls by concealing their objectives and distributing their work across separate sessions. The company said the group had also developed an offline simulation toolkit that did not depend on Claude, meaning account suspension alone could not eliminate all of the capabilities already developed.

The incident comes as AI companies increasingly rely on threat intelligence, automated classifiers and account-level enforcement to identify misuse. Anthropic said it has incorporated findings from its investigations into its safeguards and recently introduced additional classifiers designed to detect activity related to weapons development.

The immediate policy question markets is whether AI providers can detect coordinated misuse early enough to prevent harmful projects from progressing beyond experimentation, particularly when users can divide sensitive work across multiple sessions or continue development with offline tools.

Writer: Victoria Hale
Technology & Blockchain Writer

Victoria Hale writes about blockchain technology, digital infrastructure, and the intersection of emerging technologies with finance. Her articles explore how new protocols and systems are shaping the evolving digital economy.

She prioritises clarity and accuracy when explaining technical developments to a general audience.