Pentagon probe: AI, faulty intel, outdated imagery led to Iran school strike killing 123

Over 150 civilians, including 123 children, were killed in the strike on the elementary school in Minab.
The building had once been part of a military compound. Then it became a school.
The Pentagon investigation found the school's conversion from military site to civilian facility was not reflected in the intelligence used for targeting.
Mark

So the Pentagon knew this was a school, and they hit it anyway?

Mimi

Not exactly. The investigation found that the building had been part of a military compound originally. It was later walled off and converted to a school. But the people doing the targeting didn't have current information about that conversion. The intelligence was outdated.

Luke

But someone must have known it was a school by February. It was operating openly as a school. How does that information not make it into the targeting process?

Mimi

That's what the investigation is trying to answer. They found cuts to civilian-protection personnel—fewer people whose job it was to catch exactly this kind of mistake. And the system was leaning heavily on AI to process targets. More automation, fewer people checking.

Mark

What about the AI itself? Did it fail?

Mimi

Palantir, the company that built the software, said there's no evidence the software was at fault. They said the problem was the underlying data—the intelligence going in was bad.

Luke

But that's a convenient position for them. They built a tool that was supposed to help avoid this. After the fact, they built new tools to catch these mistakes. That suggests the old system had blind spots.

Mimi

True. And the UN investigation found reasonable grounds to believe it was a war crime—an indiscriminate attack. But that report isn't binding. It's evidence that could be used later, but it's not a judgment.

Mark

So what happens now?

Luke

That's the real question. The evidence exists. Whether anyone is ever held accountable is a different matter entirely.

Mimi

For now, it's a record. One hundred twenty-three children dead, and a documented failure in how the military identified and struck the target.

  • Over 150 civilians, including 123 children, were killed in minutes when two Tomahawk missiles struck a functioning elementary school — a site the targeting chain did not know had ceased to be a military compound.
  • Pentagon personnel knew within hours that American forces were responsible, yet the strike unfolded inside a 24-hour campaign that hit more than a thousand targets, its scale compressing the time and attention available for human judgment.
  • Investigators found that cuts to civilian-protection staff had hollowed out the institutional knowledge needed to catch the site's transformation, while AI systems processed and prioritized targets using imagery that no longer reflected reality.
  • Palantir, whose software was part of the targeting infrastructure, denied responsibility for the underlying data failures — then quietly developed new tools to flag the very kinds of inconsistencies that human reviewers had missed.
  • A UN fact-finding mission concluded there are reasonable grounds to believe the United States committed war crimes, and while the report carries no binding legal force, it enters the international record as evidence that could one day support accountability proceedings.

On February 28, 2026, two American cruise missiles struck an elementary school in Minab, Iran, killing more than 150 civilians — 123 of them children — in what has become one of the gravest incidents of civilian harm in recent American military history. A Pentagon investigation has since traced the catastrophe to a convergence of institutional failures: an over-reliance on artificial intelligence in the targeting process, intelligence assessments built on faulty foundations, and satellite imagery too old to reflect that a former military compound had long since become a school. The tragedy raises questions that extend beyond any single war: what obligations accompany the delegation of lethal decisions to machines, and who bears responsibility when the systems we trust do not know what they are destroying.

In the opening hours of American armed conflict with Iran, two Tomahawk cruise missiles struck Shajarah Tayyebeh Elementary School in the city of Minab. More than 150 people died — the majority of them children. It would become one of the deadliest incidents of civilian harm caused by US military operations in decades.

A Pentagon investigation has now reconstructed how it happened. The targeting process leaned heavily on artificial intelligence, drew on flawed intelligence assessments, and relied on imagery that was no longer current. The school had once been part of a military compound; at some point it was walled off and began operating openly as a place of learning. No one in the targeting chain knew. Cuts to civilian-protection personnel had left the kind of institutional gap that allows a school to remain, in a database, a threat.

Palantir, whose software was embedded in the targeting process, stated that the fault lay not in its tools but in the data and human decisions surrounding them. Yet in the aftermath, the company developed new software designed to surface the inconsistencies and disqualifying factors that reviewers had failed to catch — an implicit acknowledgment that the human layer of oversight had broken down.

The incident drew international scrutiny. A UN fact-finding mission examined the Minab strike alongside a second American attack on a sports complex and residential area in Lamerd, which killed twenty-two people and scattered tungsten pellets across a neighborhood. The mission's report concluded there were reasonable grounds to believe the United States had committed war crimes through indiscriminate attacks on civilian sites. The report is not a verdict. It carries no binding force. But it is now part of the international record — a formal accounting of how a school became a target, and how 123 children came to be counted among the costs of war.

In late February, as the United States entered armed conflict with Iran, two Tomahawk cruise missiles struck an elementary school in the city of Minab. The building was Shajarah Tayyebeh Elementary School. When the smoke cleared, more than 150 people were dead, the majority of them children. It would become one of the deadliest incidents of civilian harm caused by American military operations in recent memory.

A Pentagon investigation, reported by Bloomberg, has now documented how the strike happened. The inquiry found that the military's targeting process relied heavily on artificial intelligence, worked from faulty intelligence assessments, and used imagery that was outdated. Pentagon personnel, according to officials involved in the investigation, knew within hours that the United States was responsible for the attack. The broader context mattered: in the first twenty-four hours of the war, American forces struck more than a thousand targets. But this one—a school full of children—stood apart in its consequences.

The building itself presented a particular intelligence problem. It had once been part of a military compound. At some point it was walled off and began operating openly as a school. Pentagon investigators found that cuts to civilian-protection personnel had left gaps in the understanding of what the site actually was. No one in the targeting chain had current information that the location was now a functioning school. The reliance on AI systems to process and prioritize targets, combined with these intelligence failures and the age of the imagery being used, created a cascade of errors that ended in the deaths of 123 children.

Palantir, the company that built some of the software used in the targeting process, stated that it bore no responsibility for the underlying data or for any intelligence deficiencies in how the system was used. The company noted there was no evidence its software itself was at fault. But after the attack became public, Palantir developed new tools designed to re-examine the underlying intelligence, to identify factors that should disqualify a target, and to flag inconsistencies and inaccuracies that human reviewers might otherwise miss. The implication was clear: the human review process had failed.

The incident did not remain a matter of internal Pentagon accounting. An independent fact-finding mission, working on behalf of the United Nations Human Rights Council in Geneva, examined the strike along with a second American attack—this one on a sports complex and residential area in the city of Lamerd, which killed twenty-two people and dispersed clouds of tungsten pellets across the neighborhood. The UN team's eighteen-page report concluded that there were reasonable grounds to believe the United States had committed war crimes by launching indiscriminate attacks that resulted in civilian deaths and damage to civilian infrastructure.

The report itself carries no binding legal force. It is not a judgment. It is not a conviction. But it represents a formal finding by an international body, and it joins a growing collection of evidence that could eventually be used to pursue accountability in international courts. The question of whether that evidence will ever be acted upon remains open. For now, the investigation stands as a record of how a school became a target, and how 123 children came to be among the costs of war.

Pentagon personnel knew within hours that the US was responsible for the attack
— Officials involved in the internal Pentagon investigation
The UN team found reasonable grounds to believe the United States committed war crimes by launching indiscriminate attacks resulting in civilian deaths
— UN Human Rights Council fact-finding mission
Contáctanos FAQ