AI Likely Caused Iran School Bombing That Killed 175
Investigations point to outdated AI targeting data as the likely cause of the Minab girls' school airstrike that killed up to 180 people, most of them children.
6 min read
On February 28, a Tomahawk cruise missile struck the Shajareh Tayyebeh girls' elementary school in Minab, southeastern Iran, during school hours. Between 165 and 180 people died, most of them students aged 7 to 12. At least 95 more were injured. A second strike hit the same compound shortly after, catching first responders and parents who had rushed to the scene.
Multiple independent investigations - by NPR, CBC, The Washington Post, CNN, and The New York Times - now point to the United States as the likely source. And a growing body of evidence suggests the strike resulted from AI-assisted targeting that relied on outdated intelligence data.
TL;DR
- A US Tomahawk missile struck a girls' school in Minab, Iran on Feb 28, killing 165-180 people
- The school had been physically separated from an adjacent IRGC military compound since 2016
- Claude AI, embedded in Palantir's Maven Smart System, was used for target identification in the Iran campaign
- Pentagon refuses to confirm whether AI selected the school as a target
The School That Was Once a Military Base
The Shajareh Tayyebeh school sits next to the Sayyid al-Shuhada military complex, which houses the Asif Brigade of the IRGC Navy. Satellite imagery analyzed by multiple outlets tells a clear story of how the site changed over time.
In 2013, the school building was walled inside the IRGC compound. By September 2016, it had been physically separated - walled off with its own perimeter and converted into an all-girls elementary school. An airstrip on the complex was removed in 2024, and the surrounding land was redeveloped into housing. A clinic opened on the complex grounds in 2025. Al Jazeera reported that the school "had been separated from the military complex and had become a clearly defined civilian institution for more than 10 years."
Seven buildings were hit in total across the compound. NPR satellite analysis showed the destruction was more extensive than initially reported, with a crescent-shaped hole punched through the school's roof - damage consistent with a precision munition.
The Tomahawk Evidence
Video released by Iranian state media outlet Mehr News shows what appears to be a cruise missile striking the compound. NPR geolocated the footage to a nearby housing development under construction, confirming the location. Bellingcat independently verified it.
Jeffrey Lewis, a security expert at Middlebury College, told NPR the missile "appears consistent with a Tomahawk cruise missile." The US Navy operates these weapons exclusively among the parties involved in the conflict. General Dan Caine confirmed publicly that "the first shooters at sea were Tomahawks unleashed by the United States Navy" during the opening strikes on Iran.
President Trump offered a different account. "Based on what I've seen, I think it was done by Iran," he said. "Because they're very, inaccurate as you know, with their munitions."
Claude in the Kill Chain
The Wall Street Journal reported that the Pentagon used Anthropic's Claude AI model in planning military strikes on Iran. According to multiple sources, Claude was embedded inside Palantir's Maven Smart System, which feeds real-time satellite and surveillance intelligence into an AI that suggests targets with precise coordinates and focuses on them for commanders.
The US military struck over 1,000 targets in the first 24 hours of operations against Iran, according to Responsible Statecraft. At that volume, human review of every target was almost certainly compressed.
When Futurism asked US CENTCOM directly whether AI was used to select the school as a target, the response was: "We have nothing for you on this at this time." Defense Secretary Pete Hegseth said only that "we're certainly investigating."
No investigation has yet confirmed that Claude specifically identified the school as a military target. But the circumstantial evidence is difficult to ignore: the school appears in what was once an IRGC compound, the targeting data appears to predate the 2016 separation, and the AI systems processing that data were operating at a speed and scale that left minimal room for human verification.
The Anthropic Contradiction
The timing adds a bitter layer. Anthropic had been fighting the Pentagon over guardrails for Claude's military use. The company pushed for explicit "red lines" preventing Claude from being used for mass surveillance on Americans or fully autonomous weapons. The Pentagon demanded access for "all lawful purposes."
Just hours before the strikes began, President Trump ordered federal agencies to stop using Anthropic's technology, giving them six months to phase it out. Defense Secretary Hegseth declared Anthropic a supply chain risk. Yet multiple sources confirmed the military continued launching Claude in active operations against Iran after the ban was announced.
Anthropic CEO Dario Amodei told CBS News: "We believe that crossing those lines is contrary to American values." He added: "Disagreeing with the government is the most American thing in the world."
Human Rights Watch has called for the school attack to be investigated as a potential war crime.
What Happens Now
The EJIL (European Journal of International Law) published an analysis arguing that AI-assisted targeting creates accountability gaps that existing international humanitarian law wasn't designed to handle. When an AI system processes decade-old satellite imagery and outputs a target coordinate, who is responsible - the model's developer, the military operator, the commander who approved the strike list, or the system itself?
That question was theoretical until February 28. It isn't anymore.
A CBS News report confirmed that a preliminary US assessment found America was "likely" responsible for the strike but did not intentionally target the school. The assessment pointed to the use of outdated intelligence that wrongly identified the area as still part of an active military installation.
The site's previous coverage of Anthropic's safety pledge and the AI safety exodus from major labs takes on sharper significance now. So does the broader question raised by AI agent autonomy research: what happens when AI systems make decisions faster than humans can check them?
One hundred and seventy-five children and teachers are dead. The AI system that may have marked their school as a military target was trained to be helpful, harmless, and honest. The investigation into what went wrong has barely started.
Sources:
- Exclusive: AI Error Likely Led to Iran Girl's School Bombing - This Week in Worcester
- Video appears to show U.S. cruise missile striking Iranian school compound - NPR
- Satellite images show Iran school strike hit more buildings than earlier reported - NPR
- Evidence suggests the deadly blast at an Iranian school was likely a US airstrike - The Washington Post
- Pentagon Refuses to Say If AI Was Used to Select Elementary School as Bombing Target - Futurism
- Anthropic's Claude AI being used in Iran war by U.S. military - CBS News
- US used 'Claude' to strike over 1000 targets in first 24 hours - Responsible Statecraft
- US/Israel: Investigate Iran School Attack as a War Crime - Human Rights Watch
- United States was "likely" responsible for bombing of girls' school - CBS News
- AI and the Commission and Facilitation of International Crimes - EJIL: Talk!
