A US Defense official has disclosed that the Pentagon may use generative AI chatbots—potentially including ChatGPT and Grok—to rank and prioritize military strike targets, with humans responsible for vetting the results. This adds a conversational AI layer on top of the existing Maven system, which uses computer vision to analyze drone footage and identify targets. The disclosure comes amid scrutiny over a US missile strike on an Iranian girls school that killed over 100 children. Anthropic's Claude has reportedly been used in operations in Iran and Venezuela, though its relationship with the Pentagon is now strained after a supply chain risk designation. OpenAI and xAI have since signed agreements for classified military use of their models. The shift from traditional computer vision AI to generative AI raises concerns about output verifiability and the speed-versus-accuracy tradeoff in life-or-death targeting decisions.
Sort: