Pentagon AI Tool Failed to Flag Outdated Data Before Fatal Strike

An unreleased Pentagon investigation has found that military personnel placed excessive trust in an AI-powered targeting platform before a February strike that killed more than 150 people at an Iranian elementary school, according to officials who spoke with Bloomberg.

The strike hit Shajarah Tayyebeh Elementary School in Minab, Iran, on the opening day of conflict, killing at least 123 children. U.S. Central Command personnel relied heavily on the Maven Smart System, an AI data integration and targeting tool built by Palantir Technologies, to process and prioritize potential targets. The system recommended the school site based on outdated records that still classified the location as an Islamic Revolutionary Guard Corps facility.

Three Cascading Failures

Pentagon officials described three interconnected breakdowns in the targeting process. First, some personnel expected Maven to automatically flag stale intelligence or contradictions in target data, though it remains unclear why they believed the system had that capability. The AI compressed target-list work from hours into minutes, but did not catch the critical error.

Second, military databases had listed the Minab compound as a military site for years, even though commercial satellite imagery from 2018 clearly showed painted walls, a soccer field, assembly rows, and playground markings. Construction separating the school from an adjacent base was completed by 2017. One analyst spotted changes as early as 2019 and logged notes in a system not connected to the primary targeting database—those warnings never reached strike planners.

Third, staffing for civilian harm mitigation teams across the Defense Department had dropped roughly 90 percent in recent years, shrinking to fewer than 20 people total. Central Command's team went from 10 people to one. No mitigation team member reviewed the Minab site before the attack.

The Trump administration had demanded an overwhelming opening assault, with more than 1,000 Iranian targets hit in the first 24 hours. That operational tempo compressed the time available for additional verification.

Why It Matters

This incident exposes a critical gap between how AI targeting tools actually function and how military personnel believe they work. As defense organizations accelerate AI adoption to gain operational speed, the Minab strike demonstrates that automation can amplify—rather than catch—human errors when users assume capabilities that don't exist. The 90 percent reduction in civilian harm review staff suggests institutional safeguards were dismantled at the same time AI tools were being rapidly deployed, creating conditions for catastrophic failure.

Palantir Response and System Updates

A Palantir spokesperson told Bloomberg the company "is not responsible for the underlying data nor identifying intelligence deficiencies" and that no evidence suggests its software malfunctioned. Two sources familiar with Palantir's Pentagon contracts confirmed the government maintains primary responsibility for data quality fed into Maven.

After the strike, Palantir added features that "re-review underlying intelligence to identify factors that would disqualify a target and flag inconsistencies and inaccuracies that human review may have missed," according to a person familiar with the work. The new functionality has reportedly caught some anomalies.

A United Nations fact-finding mission this week concluded there were reasonable grounds to believe the United States committed a war crime by launching an indiscriminate attack, finding the failure to verify the target "went beyond mere negligence."

More than 120 House Democrats requested information in March about Maven's role in identifying the site. The Pentagon has not responded publicly, citing an ongoing investigation that officials say has been substantially complete for months.

These details were first reported by Bloomberg.