Anthropic's September threat report reveals a Yemen-based cell assessed as linked to the Houthis used Claude Code to develop guidance software for guided rockets, ballistic missiles, and hypersonic glide vehicles by operating multiple parallel instances for coding, research, and review. The operators circumvented safeguards through task fragmentation, test-fired a weapon, conducted failure analysis using telemetry, and compiled an offline toolkit before accounts were disrupted.
Anthropic reported that military and government actors in Iran, Yemen, and the UAE used Claude AI for weapons development, targeting, and influence campaigns between December 2025 and August 2026. Yemen-based actors used Claude to develop guidance software for ballistic missiles and conducted test-fires, while Iranian actors compiled targeting data on US Navy vessels and conducted surveillance operations. Anthropic disrupted the malicious activity, banned accounts, and shared threat intelligence with partners.
Anthropic reports blocking multiple malicious uses of its Claude AI model, including attempts to develop missile-guidance software in Yemen, Russian cyber-espionage targeting Ukraine and Europe, Chinese operations against Middle Eastern and Asian networks, and Iranian influence campaigns. The company says it banned associated accounts and shared threat intelligence with partners, though some malicious requests evaded safeguards by obscuring intent across separate sessions.
Anthropic's September 2026 report details rogue actors using Claude AI to develop weapons, including Houthis building ballistic and hypersonic missiles, Russian groups creating autonomous killer drones, and China-based actors designing fire-control systems. The report documents six cases of serious weaponization attempts across Yemen, Russia, and China, with some bioweapon development efforts partially bypassing Anthropic's safety controls.