A Pentagon internal review, reported by Bloomberg, concluded that overreliance on AI tools helped contribute to a Feb. 28 missile strike that killed more than 150 people, including at least 123 children. The Palantir system Maven, described as "AI-enabled," fuses multiple intelligence sources but was not designed to flag outdated or inconsistent data. The review called the technological trust one of several preventable errors and noted that staffing cuts left no one to review the strike plan.
Pentagon Review: Overreliance on Palantir AI Contributed to Deadly Iran School Strike

An internal Pentagon review, reported by Bloomberg, found that excessive trust in fallible artificial intelligence tools was a contributing factor in a Feb. 28 missile strike on an Iranian school that killed more than 150 people, including at least 123 children.
What the Review Found
The investigation identified reliance on a Palantir-made system called Maven as one of several preventable errors. Described in the review as "AI-enabled software," Maven aggregates and fuses multiple streams of intelligence to support commanders' targeting decisions. In the rushed and chaotic buildup to the Iran war, some personnel assumed the tool would flag outdated or inconsistent data — capabilities the system was not trained to provide, the report said.
Process Failures and Staffing Cuts
The Department of Defense has increasingly positioned Maven as a central part of its intelligence toolkit over the past year. At the same time, the report noted that staffing reductions under the previous administration sharply reduced teams tasked specifically with minimizing civilian harm. Bloomberg reported that none of the remaining personnel on those teams reviewed the strike plans before the school was attacked.
Bloomberg’s account cites Pentagon officials involved in the internal review and describes technological trust as one element in a chain of avoidable errors that led to tragic civilian losses.
For full coverage, see the Bloomberg report: www.bloomberg.com.
Help us improve.























