Anthropic Blocks Use of Claude AI Model in Weapon Software Development
AI lab Anthropic blocked attempts by actors linked to Yemen, Russia, and China to use the Claude model for missile and drone software.
Anthropic announced that its Claude artificial intelligence model was attempted to be misused for developing software for missiles, guided rockets, underwater warfare, and autonomous drone swarms.
Yemen-Linked Weapon Attempts
Anthropic announced that it detected and blocked the activities of a Yemen-based weapons engineering cell between December 2025 and August 2026.
It was stated that these actors used Claude Code to guide, navigate, and stabilize a flying vehicle, and consulted the AI again after a tested guided rocket failed.
China and Russia Operations
In a China-linked operation, it was reported that Claude was used to prepare fire control specifications for underwater warfare and to work on anti-torpedo systems.
In another Russia-based operation, it was stated that codes to manage attack and surveillance behaviors for autonomous military drone swarms were written and tested.