On the morning of February 28th, the first day of Donald Trump’s war on Iran, the children of the Shajareh Tayyebeh elementary school in the southern Iranian town of Minab were at their desks a little after ten o’clock when the missiles arrived. The name of the school means “The Good Tree.”
By the time the dust settled, as many as 175 people were dead, most of them girls between the ages of seven and twelve. Iran’s confirmed count came to around 155, and the list its state broadcaster released documents dozens of dead little boys, dozens of dead little girls, more than two dozen dead teachers, several dead parents, a dead school bus driver, and a dead pharmacy technician from the clinic next door.
The teachers had started phoning families the instant the attacks began, begging them to come collect their children. There wasn’t enough time. Some parents reached the school only in time to claw through the rubble looking for their daughters, and according to first responders and a Reuters stringer the building was hit a second time, a so-called “double tap” (which is a war crime), with the survivors and the rescuers caught in that second blast and blown to pieces.
Amnesty International and Human Rights Watch both opened investigations. Reporters at TIME and elsewhere traced the weapon to an American Tomahawk cruise missile, the kind fired only by US forces in that war, and a preliminary military inquiry concluded our forces were almost certainly responsible for both the initial killing of the children and the double-tap that killed the firefighters, rescuers, and their parents.
When a reporter asked Trump about it at the G7 summit in Évian-les-Bains last week, he waved the dead children away as if they were nothing. “Nobody did that on purpose,” he said. “Mistakes are made, war is nasty.” Then he handed the question off to Pete Hegseth and moved on.
His shrug doesn’t survive contact with the evidence. An Al Jazeera investigation that traced satellite imagery back more than a decade found the school had been a clearly marked civilian building, walled off from the neighboring military base with its own separate gates since 2016.
A medical clinic that opened on the same complex barely a year before the strike was left untouched, while the school full of children was hit.
Either our targeting was precise enough to spare a brand-new clinic but careless enough to level a decade-old school, or the school was struck on purpose. The children, after all, were the kids of members of Iran’s military. Both answers are damning, along with the double-tap evidence, and there’s exactly one document that would tell us which is true.
That document exists. The Pentagon finished its investigation last month, and members of Congress still haven’t been allowed to see it. So a bipartisan group of senators has resorted to the only leverage they have left, moving to freeze 75 percent of Hegseth’s travel budget through the defense authorization bill until he hands over the unredacted civilian-harm investigations, Minab among them.
The same provision demands the unedited video of the boat strikes off Venezuela that have killed more than 200 people, which Trump and Whiskey Pete are also refusing to release. You don’t bury a report that clears you.
We’ve seen this reflex before, from the Amiriyah shelter in Baghdad in 1991, where American smart bombs killed more than 400 civilians the Pentagon insisted were a military target, to the Doctors Without Borders hospital in Kunduz in 2015, where the official story changed three times before a US commander finally admitted the decision had been entirely American. Not to mention My Lai in Vietnam. Deny, deflect, and bury, every single time.
There’s a reason, however, that this particular cover-up may be more frantic than the ones that came before it, and it carries a name we’ve all heard way too many times: the machine helping choose our targets in Iran was Grok, Elon Musk’s AI chatbot.
We know this because the Pentagon’s chief digital and artificial intelligence officer, Cameron Stanley, said so under oath, swearing in a court filing that Musk’s “Grok Gov Model” let American forces “deploy over 2,000 munitions to 2,000 distinct targets within 96 hours” during the operation. That’s one target every three minutes.
Musk’s company signed its deal with the Defense Department on February 23rd. The war began February 28th. And we only learned any of this because the disclosure slipped out in a lawsuit the NAACP brought over a Musk data center accused of poisoning a Black neighborhood near Memphis with gas-turbine fumes.
This is the same Grok that spent the summer of 2025 praising Adolf Hitler, calling itself “MechaHitler,” telling users that people with Jewish surnames were the problem and that the Holocaust was the solution, and injecting white-supremacist “white genocide” talking points into conversations that had nothing to do with them.
That is the AI a Nazi-salute-giving apartheid-era South African billionaire wired into our targeting systems five days before we started killing Iranian schoolchildren.
I can’t tell you that Grok chose the school in Minab, and neither can anyone else, because the report that would answer that question is the one they refuse to release. That’s the point.
An AI that has openly admired a genocidal Hitler was helping aim American missiles at brown people at the rate of one target every three minutes, the deadliest day for civilians in the entire war. It happened on the first full day that machine was running, and the people who could tell us what it did are betting we’ll lose interest before they ever have to reveal what really happened.
