How Palantir's Maven Smart System compressed the kill chain — and why 175 people, most of them children, paid the price.
On February 28th, Palantir's Maven Smart System did its job.
Using Claude, it processed intelligence data from the Department of Defense, generated coordinates for each strike, ranked targets in terms of strategic importance, and even "draft[ed] automated legal justifications for each strike." With Maven, the U.S. military was able to hit more than 1,000 targets in the first 24 hours of the war.
One of those "strategically important" targets
Minab, Iran
Source: New York Times
The AI-powered Maven Smart System compressed the kill chain so dramatically that human oversight failed to keep up. As a result, over 100 innocent schoolchildren — each with big dreams and bright futures — are dead.
"Had that momentum kept going, there would have been, at the very least, more people in place focused just on this problem, double and triple-checking that every single entity we were looking at targeting in Iran was actually a valid military target."
— Wes Bryant, Air Force Special Operations veteran, former civilian harm assessor at the Civilian Protection Center of Excellence, speaking to WNYC's On the Media after Minab
The office that existed specifically to prevent strikes like Minab was already being dismantled when Minab happened.
The Civilian Protection Center of Excellence was created by Congress in 2022 to train commanders in how to avoid killing noncombatants. By early 2025, Defense Secretary Pete Hegseth had begun unwinding it. By June 2025, the Army had absorbed the Center into other offices — effectively killing it, in the words of one official.
Wes Bryant, an Air Force Special Operations veteran who had run civilian harm assessments at the Center, was placed on leave that March, resigned in September, and has since become a public critic of the administration.
Bryant called Minab "pure negligence" — a predictable consequence of dismantling institutions meant to minimize civilian death tolls.
The Center was not the only oversight casualty. According to the Brennan Center for Justice, Hegseth reduced staffing at the Office of the Director of Operational Test and Evaluation by half — the office responsible for testing 268 weapons systems, some of them with deep AI integration.
weapons systems under the oversight of an office whose staff was cut in half. Some with deep AI integration. All now less watched.
Nine days after Minab, the Pentagon "formalize[d] Palantir's Maven AI as a core military system with multi-year funding." Maven is now a "program of record" locked into the protected defense budget.
The direction was already clear before Minab. In June 2025, Vice Admiral Frank Whitworth, then director of the National Geospatial-Intelligence Agency, told an industry audience that NGA had begun producing entire intelligence products without analyst involvement. The agency adopted a disclosure label — "Machine-generated GEOINT" — to flag them. For some products, Whitworth said, "no human hands actually participate."
[S]uch AI-generated intelligence products are now being circulated at the highest levels of the US government.
— Breaking Defense
The high-level doctrine behind these decisions was written down plainly in January. In a strategy memo, Hegseth declared the Pentagon an "AI-first warfighting force" and stated the principle plainly:
"We must accept that the risks of not moving fast enough outweigh the risks of imperfect alignment."
— Defense Secretary Pete Hegseth, strategy memo, January 2025
Minab is what 'imperfect alignment' looks like in practice.
It looks like a school — its name meaning Good Tree in Farsi — reduced to rubble by coordinates generated by a machine, justified by a machine, authorized without adequate human review, after the humans who would have checked that work had been systematically removed from their posts.
The machine did its job. That was the problem.