Which AIs is the Pentagon actually putting to use?

Publicly discussed military AI work spans analysis, logistics, maintenance, cyber defense, sensing, and unmanned systems rather than one all-purpose “war AI.”…

Publicly discussed military AI work spans analysis, logistics, maintenance, cyber defense, sensing, and unmanned systems rather than one all-purpose “war AI.” The Air Force request for rapidly deployable drone swarms for base security shows the direction: networks of comparatively small systems can watch a perimeter and coordinate, while people remain responsible for rules, authorization, and accountability. Coverage should distinguish a proposal or pilot from a fielded capability and avoid turning public reporting into operational guidance.

Why the Pentagon uses many focused AIs, not one

Different jobs fail in different ways. Sorting maintenance records, watching a perimeter, planning deliveries, and spotting network attacks each need different data and different checks. A smaller system built for one job is easier to test, update, and switch off when it misbehaves, and it keeps responsibility clearer: people set the rules and answer for the decisions.

What to do, step by step

1. Ask which stage it is at: proposal, pilot, test, or regular use — those are not the same thing.
2. Ask who stays in charge: who sets the rules, reviews the output, and can stop it.
3. Look for testing results, known mistakes, and how problems get reported and fixed.
4. Separate support jobs like logistics from decisions affecting people, which need stronger human oversight.