Story perspectives

Houthi-linked cell bypassed Claude to build missile guidance

By Drooid · Sep 12, 2026, 2:50 PM UTC · How we work

Summary

Houthi Cell Uses AI

A Houthi-linked cell used Anthropic’s Claude AI to create missile guidance software.
The cell pursued a guided rocket, a >2,000 km ballistic missile, and an R2000 glide variant.
The cell bypassed Anthropic safeguards to write navigation code and then test-fired a guided rocket that failed.
Anthropic banned the accounts, shared threat intel with partners, and upgraded its detection tools.

Summary

Pentagon Blacklists Anthropic

Anthropic researcher Jacob Coxon resigned, warning that AI could kill humans within a decade.
The U.S. Pentagon blacklisted Anthropic as a supply-chain risk.
Pentagon units deployed Claude models in operations in Iran and Venezuela.
Anthropic hired an independent firm to audit recurring AI misuse incidents.
John Thickstun warned companies lack democratic oversight to decide safe versus unsafe AI behavior.

History

Historical Context

In June 2010 the United-States and Israel launched Operation Olympic Games, releasing the Stuxnet worm that infected over 200,000 computers and sabotaged roughly 1,000 Iranian centrifuges at Natanz by hijacking Siemens PLCs. That precedent of software-driven kinetic attacks now resurfaces as Houthi-linked fighters use Anthropic’s Claude AI to draft missile-guidance code, prompting Pentagon blacklists and calls for AI oversight.

1 of 3

Houthi Cell Uses AI

A Houthi-linked cell used Anthropic’s Claude AI to create missile guidance software.
The cell pursued a guided rocket, a >2,000 km ballistic missile, and an R2000 glide variant.
The cell bypassed Anthropic safeguards to write navigation code and then test-fired a guided rocket that failed.
Anthropic banned the accounts, shared threat intel with partners, and upgraded its detection tools.

1 / 3