A US intelligence analyst spotted changes at an Iranian site years before American missiles hit it and killed more than 120 children. The warning was recorded in 2019 but never reached the military database used to prepare targets. A Pentagon investigation has now exposed how outdated intelligence, rushed targeting and reliance on AI preceded the Minab school strike.
The revelation changes the central question surrounding the February 28 attack on Shajareh Tayyebeh Primary School in Minab. The problem was not simply that the United States had old information.
Someone had apparently noticed that the site had changed years earlier.
Where Palantir’s Maven entered the targeting process
A US intelligence analyst noticed changes at the site in 2019 and entered observations into a digital intelligence system, according to officials familiar with the Pentagon investigation. More than 1,000 targets were hit during the opening 24 hours of the assault. On February 28, two Tomahawk cruise missiles struck Shajareh Tayyebeh Primary School and its grounds in Minab, according to reporting based on the Pentagon investigation. More than 150 people were killed. The Pentagon investigation cited in recent reporting puts the figure at 123. The UN mission said there were reasonable grounds to believe US strikes on Minab and Lamerd on February 28 amounted to war crimes.
The information therefore did not make its way into the targeting process. The building was reviewed again in later years without the main targeting database being updated, according to reporting on the investigation. The reported number of children killed varies slightly between investigations and official accounts. The Pentagon has not publicly released the full findings of its investigation.
Some US Central Command personnel relied too heavily on Maven and expected the system to identify old or conflicting intelligence, according to officials familiar with the investigation. Palantir said it “is not responsible for the underlying data nor identifying intelligence deficiencies” and said there was no indication that its software was at fault. The company later developed tools designed to recheck underlying intelligence and flag information that could disqualify a target, according to reporting on the investigation. Officials said the pace reduced the time available to complete checks before targets were approved. Some Pentagon personnel knew within hours that the United States was responsible for the strike, according to officials familiar with the investigation.
There was one major problem. That system was not connected to the main military intelligence database used to develop strike targets. The Pentagon investigation also examined the military’s use of Palantir Technologies’ Maven Smart System. Maven brings information from a large number of military and intelligence sources into one system and helps personnel process possible targets much faster. There is no evidence, however, that Maven malfunctioned or independently decided to attack the school. The speed of the US campaign is another part of the investigation. But speeding up the process could not correct information that had never reached the right database. The Pentagon disclosures have emerged alongside a separate investigation by a UN-backed fact-finding mission. Investigators concluded that the Minab school was clearly identifiable as a civilian building and that the United States had failed to adequately verify that its targets were military objectives. The White House rejected the UN body’s findings. A UN fact-finding mission is an investigative body and does not itself decide criminal responsibility in a court.
That did not happen before the Minab strike. The military was using AI-assisted systems to process information at a speed that would previously have taken analysts much longer.
But the area changed long before the 2026 attack. Commercial satellite imagery showed walls and entrances separating the school from the military compound by 2017. Images from 2018 showed features associated with a school, including brightly painted walls, a football pitch, assembly rows and children’s play markings.
By the time the United States began its air campaign against Iran, military planners were therefore working with information that did not fully reflect what had happened at the site. That information did not reach the people who later helped decide what to strike. The Minab case leaves a more difficult question for the US military-if AI allows a target to move through the system in minutes, what happens when the information entering that system is wrong?
The site had previously been identified by the United States as an Iranian military facility.
