Pentagon Report: Overreliance on Palantir Maven System Caused Errant Strike on Iranian School, Killing 123 Children

On February 28, 2026, two U.S. Tomahawk missiles struck the Shajarah Tayyebeh primary school in Minab, Iran, killing at least 123 children.

Reconstruction of Events

According to Bloomberg, a Pentagon investigation concluded that the incident occurred on the first day of the U.S.-Israeli war against Iran. At the time, more than 1,000 Iranian targets were struck within 24 hours, and the target-processing workflow was drastically compressed. The school is adjacent to an Islamic Revolutionary Guard Corps Navy facility, and the original intelligence database still marked it as a military target, even though satellite imagery had already shown school features such as colorful walls and a soccer field. At least one analyst had noticed the change as early as 2019, but the relevant information was not synchronized to the main intelligence database.

The investigation also noted that the civilian-harm mitigation team had been significantly reduced before the incident. CENTCOM's civilian-harm team was cut from 10 people to 1, and that team was not involved in the review of the Minab target.

Mechanism Breakdown

The Maven Smart System integrates more than 150 data sources to organize intelligence, track operations, and assist commanders in target selection. The investigation found that some U.S. Central Command personnel expected the system to automatically detect outdated intelligence or inconsistencies, but the system in practice only processes the underlying data provided by the government.

Palantir said that the company is not responsible for the accuracy of the underlying data, nor does it bear responsibility for identifying intelligence deficiencies. After the incident, the company upgraded Maven's capabilities so that it can re-review intelligence to flag potentially invalid targets.

Industry Impact

This incident has become the most consequential accountability case in the field of military AI applications. A report released in May by the Department of Defense Inspector General showed that personnel cuts starting in 2025 had stalled civilian-harm mitigation programs. The incident exposed the limitations of AI-assisted decision-making in high-tempo combat environments: when the human review step is compressed, the system cannot compensate for defects in the original data.

Strategic Assessment

(The following is analysis rather than fact.) Based on existing materials, the role of AI systems in target screening is shifting from assistance to de facto dominance, but accountability remains within a traditional framework. This may force defense contractors to redesign the human-machine division-of-labor interface, while also pushing Congress to demand stricter civilian-protection review mechanisms. Similar historical incidents show that unless data synchronization and staffing problems are addressed, similar risks will persist.