The US military may use generative AI chatbots to rank lists of targets and recommend which to strike first, according to a Defense Department official speaking on background.
Under the scenario described, a list of potential targets would be fed into a generative AI system cleared for classified settings. The system would analyze the data and prioritize targets while factoring in variables like current aircraft positions. Humans would then check and evaluate the output before any decisions are made.
The official described this as an illustrative example and would not confirm or deny whether it reflects current operational practice.
Two Different AI Technologies
The military has run a separate “big data” initiative called Maven since at least 2017. It uses computer vision to process drone footage and other imagery, algorithmically flagging potential targets on a battlefield map. A 2024 Georgetown University report found soldiers using the system to select and vet targets, with friendly forces and potential targets displayed in distinct colors on an interface dashboard.
Generative AI is now being layered on top of that foundation as a conversational interface — one the military may use to surface and analyze data faster when making prioritization decisions.
The two technologies carry different risks. Maven’s map-based interface required users to directly inspect data. Generative AI outputs, built on large language models, are easier to access but harder to verify. The official noted the technology is reducing time in the targeting process but offered no specifics on net speed gains when factoring in the human review requirement.
Which Models Are Involved
Anthropic‘s Claude was the first model approved by the Pentagon for classified use. Reports from other outlets indicate it has been integrated into military AI systems and used in operations in Iran and Venezuela. OpenAI‘s ChatGPT and xAI‘s Grok have also reached agreements with the Pentagon for classified deployment and could, in theory, be used for targeting scenarios in the future.
For non-classified tasks — drafting presentations, analyzing contracts — the Pentagon began offering generative AI access to millions of service members in December through a program called GenAI.mil.
The disclosure arrives as the Pentagon faces scrutiny over a strike on a girls’ school in Iran in which more than 100 children died. Multiple outlets have reported the strike involved a US missile; the Pentagon says the incident is still under investigation. A preliminary inquiry, according to the report, found outdated targeting data to be partly responsible. No evidence has yet emerged explaining what role, if any, generative AI systems played in that specific strike.
Photo by Pixabay
This article is a curated summary based on third-party sources. Source: Read the original article
