A defense official reveals how AI chatbots could be used for targeting decisions

A Defense Department official has disclosed to MIT Technology Review that the U.S. military is exploring the use of generative AI chatbots, including models from Anthropic (Claude), OpenAI (ChatGPT), and xAI (Grok), to analyze data and rank potential targets for strikes. According to the official, who spoke on background, these systems could process lists of targets and provide prioritized recommendations by factoring in dynamic battlefield conditions like aircraft positioning. Crucially, human operators would retain final decision-making authority, responsible for vetting and evaluating the AI's outputs before any action is taken. This revelation provides new insight into the specific, accelerating role conversational AI may play in military operations.

The disclosure arrives as the Pentagon faces intense scrutiny over a recent missile strike on a girls' school in Iran, which is under investigation. While there is no evidence directly linking generative AI to that incident, reports indicate Anthropic's Claude has already been integrated into existing systems like Project Maven and used in operations in Iran and Venezuela. Project Maven, a 'big data' initiative active since 2017, relies on older computer vision AI to identify targets from imagery. The new generative AI layer represents a fundamental technological shift, offering a faster, conversational interface to query and analyze data, but one that produces outputs that are easier to access yet harder to verify than Maven's map-based system. The Pentagon's GenAI.mil program has expanded AI access, but only a few models are approved for classified use, marking a significant escalation in deploying less battle-tested LLMs for critical defense functions.