Anthropic’s September 2026 threat intelligence report documents six conventional-weapons investigations across Yemen, China, and Russia. Case GTG-87001 describes a cell based in northern Yemen running three weapons development programs: a guided rocket using a commodity phone-class flight computer with terminal homing guidance; a multi-stage ballistic missile with a stated range goal above 2,000 km; and an “R2000” missile set including a hypersonic glide vehicle variant.
The actors used Claude Code in place of human software engineers to develop guidance, navigation, and control (GNC) software. They managed several Claude instances at once with assigned roles — one writing code, one researching, one reviewing — much as a lead would delegate on a small engineering team. Claude supported autopilot integration, flight-control and position-estimation code, tuning, firmware build pipelines, and flight simulation.
Anthropic disrupted the activity between December 2025 and August 2026. The actors test-fired a guided rocket in Yemen; the test appears to have failed, and within hours they returned to Claude to diagnose the failure. Anthropic found no evidence the actors succeeded in fielding an operational device. The group compiled an offline engineering toolkit before accounts were banned.
Anthropic uses internal designators called Generative Threat Groups (GTGs) for actors abusing AI. The report measures “uplift” — speed, scale, and depth of harm enabled by AI versus without AI.