Amodei's Iran School Strike Remark Backlash

Editor J
Amodei's Iran School Strike Remark Backlash

Amodei called the Iran school strike that killed 120 children a case that 'doesn't violate our red lines' — even as Claude powers the Maven targeting system.

Comments by Anthropic CEO Dario Amodei have sparked intense backlash. During a Bloomberg interview with Emily Chang regarding a U.S. airstrike on a primary school in Minab, Iran—which killed over 150 people, including around 120 children—Amodei called the incident 'a really terrible thing to happen,' but added that it was 'a use case that doesn't even violate our red lines.'

The controversy deepened as Dario Amodei addressed reports of Claude's military use in the targeting chain behind the Iran school strike. Admitting that Anthropic lacks visibility into the model's exact role—'We don't have access to, we don't know exactly how these models were used'—he still asserted that 'the principle that a human makes the final decision was obeyed.' Critics asked how he could guarantee the principle held without knowing how the model was used.

The Reality of the Iran School Strike

A look at the Iran school strike itself challenges the claim of a principled operation. Per Amnesty International, the Shajareh Tayyebeh Elementary School in Minab was directly hit on the morning of February 28, the first day of U.S. operations against Iran. Classes were in session, and satellite analysis showed the school's roof pancaked by a top-down blast.

The Iran school strike left over 150 dead, including 110 to 120 children, by Iranian authorities' count. Amnesty International said an internet blackout hindered independent verification, but concluded that a U.S. Tomahawk cruise missile—used exclusively by the U.S. in this war—was the likely weapon.

The school shared a wall with an Islamic Revolutionary Guard Corps (IRGC) compound, yet satellite records show it had been split off years earlier into a clearly marked civilian school, playground lines and all. The Pentagon has not claimed responsibility for the Iran school strike, though Defense Secretary Pete Hegseth promised a 'thorough probe' in March—read by The Washington Post as a tacit admission.

Claude's Integration into the Targeting Chain

Claude's role sits inside the Maven Smart System, an AI targeting platform Palantir built under a $1.3 billion Pentagon contract for U.S. Central Command. The Maven Smart System fuses satellite, drone, and signals intelligence onto one screen, generating, ranking, and pairing targets to weapons. Claude's military use is wired into its natural-language and analysis layer.

A hand operating a laptop in a military operations setting
Operating a targeting system on a computer in a military setting

Scale is the point. Through the Maven Smart System, CENTCOM struck 1,000 targets in the first 24 hours against Iran and 13,000 over the 38-day Operation Epic Fury. By the Pentagon's own data on Claude's military use, peak daily AI token usage spiked as much as 44-fold.

The Maven Smart System is built to compress the time from detection to strike. Hamza Chaudhry of the Future of Life Institute warned that targeting this fast can reduce human sign-off to a mere rubber stamp. Even Jack Shanahan, the retired general who built Maven, asked: 'You may have a thousand targets, but are they the right targets?'

The Limits of Anthropic's Red Lines

Picking the right targets runs straight into Dario Amodei's claim that human oversight held. Earlier this year, Anthropic clashed with the Trump administration by barring Claude's military use in fully autonomous weapons and mass domestic surveillance. In response, the Pentagon branded Anthropic a supply-chain risk, and the company sued. These are the Anthropic red lines now under scrutiny.

This fight is not new. The same Anthropic red lines over Claude's military use surfaced in the company's court battle with the Pentagon over NSA cyber operations. The contradiction is stark: a company that drew a firm line against autonomous weapons and sued the federal government still has its AI running inside a targeting platform that hit a school full of children.

Whether Claude touched that specific Iran school strike remains unknown. What is certain is that it was part of the system that produced it. The moment Dario Amodei's Anthropic says it cannot see how the model was used, its claim that the Anthropic red lines held wobbles right along with it.

List Next ›
Menu