Pentagon Ties Palantir's Maven AI to Iran School Strike
A.I. / news
Pentagon Ties Palantir's Maven AI to Iran School Strike
Investigators found Centcom's civilian harm team had fallen from 10 people to one before the Feb. 28 strike killed at least 123 children, Gizmodo and IBTimes UK reported.

Pentagon investigators found that Central Command's civilian harm mitigation team had shrunk from 10 people to one in the months before a Feb. 28 missile strike on an elementary school in Minab, Iran, killed at least 123 children, according to an unreleased internal Pentagon review described independently by Gizmodo and IBTimes UK.
Two Tomahawk missiles hit Shajarah Tayyebeh Elementary School on the opening day of the Iran war, killing more than 150 people in total. Investigators concluded that personnel at Centcom relied heavily on Maven Smart System, an artificial-intelligence targeting platform built by Palantir Technologies that fuses satellite imagery, drone feeds and signals intelligence, and that the system pulled the school out of a batch of candidate targets and returned it as a recommended day-one strike.
Why the school was on the list
U.S. intelligence databases had the Minab compound catalogued as a facility linked to Iran's IRGC, the country's paramilitary Guard Corps, for years. IBTimes UK reported that an intelligence analyst had flagged changes at the site as early as 2019, but that the flag was entered into a system not connected to the one strike planners used, so it never reached them. Work that normally took hours of human review was compressed into minutes using Maven, according to Gizmodo's account of the review.
A staffing collapse across the department
The civilian harm mitigation program lost roughly 90 percent of its staff department-wide in the years before the war, shrinking to fewer than 20 people across the entire U.S. military, both outlets reported, citing the same internal review. No member of a civilian harm mitigation team reviewed the Minab site before the missiles launched.
| Date | Event |
|---|---|
| Years before 2026 | Civilian harm mitigation staffing falls roughly 90 percent department-wide, to fewer than 20 people |
| 2019 | An analyst flags changes at the Minab site; the flag never reaches strike planners, IBTimes UK reported |
| Feb. 28, 2026 | Two Tomahawk missiles hit Shajarah Tayyebeh Elementary School in Minab, killing more than 150 people |
| Sept. 20, 2026 | Gizmodo and IBTimes UK report findings from the unreleased Pentagon review |

What Palantir says, and what the toll actually was
Palantir told both outlets that it "is not responsible for the underlying data nor identifying intelligence deficiencies" in the targeting failure, and that the software cannot correct inaccurate information it is given. The company has since added features that make Maven re-review underlying intelligence to flag inconsistencies human reviewers may have missed, according to Gizmodo.
The death toll itself is contested. Gizmodo's figure, drawn from the same review, is more than 150 dead including at least 123 children. Rep. Sara Jacobs, a Democrat who has reviewed material related to the strike, told IBTimes UK that Americans should feel "deeply ashamed by the killing of 175 civilians, mostly little children," a higher count than the Pentagon's own working figure. Neither outlet reported a reconciled number as of publication.
This site has previously reported on AI-linked failures in military settings, including a chatbot hallucination that nearly led armed servicemembers to board a Chinese vessel, and on Anthropic's own public call for mandatory AI kill switches in high-stakes deployments generally.
The Pentagon has not released the review publicly. Congressional inquiries into the strike remain open, and no disciplinary action has been announced, IBTimes UK reported.
Sources
More in A.I.
- 01GPT-Rosalind-6 Reuses GPT-6 Astra's Weights, Cuts Biology Refusals to 9.4% and Stays Behind a GateOpenAI's Oct. 30 system card addendum says no training was added, so the difference from Astra is deployment controls and who is allowed in.
- 02Architect Says Claude Opus 5.5 Drew a House Set Good Enough for a Permit After Two Redline RoundsStephen Gallagher's informal test of 13 models gave Opus 5.5 the most human feedback, and the drawings still missed a required second exit.
- 0350 GPT Agents in Minecraft Built a Ration Ledger, and 34 Died Doing ItY Combinator's Labs group and Olam Labs ran GPT-6.1 Sol and Claude Opus 5.5 swarms through Minecraft and Mindustry. GPT won both head-to-head games, but each result comes from one harness and one setup.
- 04China's Announcement No. 48 Puts 23 AI Technologies on a Draft Export ListExporters of listed technology must already register with MOFCOM. Comments on the draft catalogue close on November 15, and the sources disagree on whether the thresholds are published.