Here's a number that should keep you up tonight: 20 soldiers using the Pentagon's Maven Smart System can now identify and process as many targets as 2,000 humans working without it. That's not a typo. That's a 100x force multiplier, and it's live in Operation Epic Fury right now.

The U.S. military has hit more than 12,000 targets in its monthlong war with Iran, including over 1,000 in the first 24 hours after operations launched on February 28. Admiral Brad Cooper, the CENTCOM commander, confirmed in a March video update that soldiers are "leveraging a variety of advanced AI tools" in the campaign. The official line: humans always make the final call on what to shoot.

The unofficial reality is more complicated.

One of the sites bombed on day one was an Iranian school. At least 175 people died, most of them children. More than a hundred lawmakers in both the House and Senate have signed letters to Pentagon chief Pete Hegseth demanding to know whether Maven was involved in selecting that target. At a closed-door House Armed Services Committee briefing on March 25, officials told lawmakers AI handled "data management" but not final target selection. That distinction is doing a lot of heavy lifting.

"For somebody who spent years talking about how we're moving too slow, I'm now concerned about how fast we're moving," said Jack Shanahan, the retired lieutenant general who led the military's original AI integration efforts. When the guy who built the system tells you he's worried, you listen.

Maven has been in development for nearly a decade. It fuses satellite imagery, intelligence feeds, asset movements, and sensor data into a single platform that can process targets in seconds instead of hours. The early phase of the Iran campaign relied on pre-existing Pentagon target lists. But as those run out, AI is taking a bigger role in finding new ones.

"We are now entering the phase where those targets have been attacked and now you could potentially start to see an even greater impact of AI," said Emelia Probasco, a senior fellow at Georgetown's Center for Security and Emerging Technology. "You're looking for time-critical targets, targets that move, targets that we didn't know about before."

This is the pivot point. The AI systems that were sold as decision-support tools are becoming decision-shaping tools. When Maven surfaces a target in seconds and a human has seconds to approve it, the distinction between "recommendation" and "decision" gets very thin. Especially at 3am. Especially when 12,000 targets have already been hit and the pace shows no signs of slowing.

The Pentagon insists this is about efficiency, not autonomy. But efficiency that produces 175 dead children on day one raises a question that no briefing has answered: if the AI is just managing data, who picked that school?

First reported by USA Today, with additional reporting from Georgetown CSET research.