Generative AI for Lethal Drone Development

Reports describe Russian-linked developers using Anthropic’s Claude to build software for autonomous FPV drones, including systems designed to identify targets and issue detonation commands without human confirmation.

Development efforts drew on Ukrainian combat footage and included testing on simulated targets or hardware; some also pursued drone-to-drone communication for swarm coordination. The accounts have raised questions about how readily available generative AI can assist weapons development and how providers can detect and limit misuse. Anthropic said it investigated the activity, terminated associated accounts, and shared findings with government and industry partners.