Anthropic CEO Addresses Possible Claude Role in Iran School Strike

Anthropic CEO Dario Amodei told Bloomberg's "The Circuit" that he does not know what role Claude played in a US military strike on February 28 that killed at least 120 children and over 150 people in total at the Shajareh Tayyebeh Elementary School near Minab, Iran. "We don't have access to, we don't know exactly how these models were used," Amodei said, per Bloomberg. He added: "The principle that we have established, and I think the principle that was obeyed here, is a human makes the final decision" -- though as The Next Web notes, that claim is difficult to verify without knowing how Claude was used. Amodei described the strike as "a really terrible thing to happen" but said the use case did not violate Anthropic's policies.

Claude is embedded in Maven Smart System, an AI-assisted targeting platform built by Palantir under a $1.3 billion Pentagon contract, per The Next Web. The system uses Claude and other AI tools to generate targets, rank them by strategic importance, and help pair weapons to targets. US Central Command struck 1,000 targets in the first 24 hours of operations against Iran and reportedly 13,000 by April 6 -- little more than five weeks into the campaign -- per The Next Web. The scale reflects Maven's design goal: compressing the time between target identification and strike.

Amnesty International reported the Shajareh Tayyebeh Elementary School was hit on the first day of US operations in Iran, killing at least 120 children and over 150 people in total. Investigations by Amnesty, Human Rights Watch, and several news organizations concluded a US-manufactured Tomahawk missile was likely used, per The Next Web. The Pentagon has not publicly claimed responsibility but is investigating; Defense Secretary Pete Hegseth promised a "thorough probe," per the Washington Post.

Retired Air Force Lieutenant General Jack Shanahan, who created Project Maven, warned that integrating Claude into Maven Smart System could lead to "unexpected impacts" and dilute human judgment, per The Next Web. "If you make more decisions rather than the right decisions, you may have a very flawed decision-making process," Shanahan said at a Stanford University workshop. Hamza Chaudhry of the Future of Life Institute separately warned that AI-accelerated targeting may compress human review to a "rubber stamp," per The Next Web, noting that 13,000 targets in five weeks supports that concern.

Amodei earlier refused to allow Claude to be used in fully autonomous weapons or mass domestic surveillance; the Pentagon responded by designating Anthropic a supply-chain risk. A federal judge blocked the Pentagon's effort to sever ties with Anthropic, but an appeals court denied a separate request for a temporary block, per The Next Web. That ongoing lawsuit sits alongside Claude's confirmed embedding in a targeting platform that generated 13,000 targets over five weeks.

Observers should follow the Pentagon investigation for attribution findings on the Minab strike; monitor Anthropic transparency reports on deployment controls; and track regulatory or legislative action on AI use in lethal-targeting decisions. For practitioners, the episode illustrates how AI outputs embedded in multi-step military decision workflows can be difficult for model providers to audit or constrain post-deployment.