Anthropic: Yemen Cell Used Claude Code for Missile Guidance
TL;DR
- Anthropic's September 2026 threat report says a northern Yemen cell used Claude Code 'in place of human software engineers' to build missile guidance software.
- The cell pursued three weapons programs, including a multi-stage ballistic missile with a stated range goal above 2,000 km and an R2000 set with a hypersonic glide vehicle variant.
- Operators split work across multiple sessions to hide intent; a guided-rocket test-fire failed, and within hours the group returned to Claude to debug it.
A cell of threat actors in northern Yemen used Claude Code 'in place of human software engineers' to develop guidance, navigation and control software for missiles, Al Jazeera reports from Anthropic's September 2026 threat report.
The Yemen file, catalogued as GTG-87001, describes three weapons programs running in parallel: a guided rocket built on a commodity phone-class flight computer with final-phase homing, a multi-stage ballistic missile with a stated range goal above 2,000 km, and a multi-variant 'R2000' set that included a hypersonic glide vehicle variant. Operators used Claude Haiku, Sonnet, Opus and Claude Code together.
They split work across multiple sessions so no single session revealed their full intent, and hid the products the software was meant for. 'The operators appeared to have conducted an unsuccessful test-fire,' Anthropic says. Within hours of the failed shot, the group returned to Claude to work out why.
The Yemen case sits alongside other operations Anthropic says it disrupted. A Russian intrusion set with the hallmarks of Midnight Blizzard used 'automated AI workflows to run nearly the entire operation, handling everything from phishing and setup to data theft' against Ukrainian, European and diplomatic targets. University students in Hunan province used Claude 'as the engineering and orchestration layer' against government and corporate networks across the Middle East, Europe and Southeast Asia. A China-aligned account ran a 'multi-day recruitment operation to infiltrate Uyghur targets in Syria,' with the model drafting outreach 'in the regional dialect' and translating replies in real time. Three Iranian state-aligned accounts ran covert influence and psychological operations tied to named propaganda institutions.
Anthropic banned the associated accounts and shared threat information with partners. Our earlier write-up covered the Russian and Chinese components of the same disclosure a day before the Yemen case surfaced in the wire coverage.
Originally reported by aljazeera.com
Read the original article →Original headline: Anthropic Threat Report Details Yemen Weapons Cell Using Claude Code as 'Human Software Engineer' for Missile Guidance