US Central Command overhauls AI targeting protocols after deadly school strike killed 156 in Iran

US Central Command has rewritten its playbook for how artificial intelligence is used to select and vet lethal targets in combat, a direct response to a February strike that killed approximately 156 people at an elementary school in Iran, including 120 children.

The Minab school strike on February 28 was the single deadliest civilian incident during Operation Epic Fury, the US military campaign against Iran. Seven months later, CENTCOM announced a series of changes designed to prevent the kind of intelligence failures that turned a school into a target.

What changed and why

The updates, reported on September 22, center on three areas: how targets get vetted before a strike is authorized, how civilian presence near potential targets is tracked in real time, and how the underlying AI systems process and refresh their data.

At the core of the overhaul sits Palantir Technologies’ Maven Smart System, the AI platform the Pentagon has increasingly relied on to accelerate the targeting cycle. The system received what CENTCOM described as extensive upgrades, including new capabilities to accurately classify building functions.

CENTCOM also deployed AI agents tasked with continuously re-evaluating intelligence, rather than relying on static snapshots that may be days or weeks old. The Minab strike exposed a critical vulnerability: the intelligence databases feeding targeting decisions had gone stale.

Open-source data feeds have been integrated into the vetting workflow as well, adding a layer of publicly available information to cross-check classified intelligence.

Speed versus safety

The updated Maven Smart System enabled the military to strike around 1,000 targets in a single day, a pace that would have been unthinkable in previous conflicts.

CENTCOM Commander Adm. Brad Cooper and other military officials have emphasized that human judgment remains essential to the targeting process, even as AI tools handle the heavy analytical lifting.

Congressional scrutiny has intensified since the strike. Senate officials have raised pointed questions about how AI-generated targeting data is managed, stored, and refreshed, and whether the speed of automated processes creates blind spots that human reviewers cannot realistically catch in the time allotted.

A broader Pentagon push

The Pentagon had already been expanding AI’s role in targeting operations prior to the strike, accelerating efforts that began gaining momentum as early as April 2026.

Palantir, the defense technology company founded by Peter Thiel, has been a primary beneficiary of this push. Its Maven Smart System, originally developed under Project Maven, has become central infrastructure for how the US identifies and prosecutes targets in the field.