One school strike in Iran has turned into a bigger fight over whether “safe” artificial intelligence can still help power deadly war decisions.

Quick Take

• Anthropic chief Dario Amodei said he does not know exactly how Claude was used in the strike.
• Reports say Claude was integrated into the Maven Smart System, which helped rank targets for U.S. Central Command.
• Investigators and outside reports point more to stale data and human verification failure than an AI mistake.
• The case has raised fresh fears about secrecy, corporate accountability, and how fast war can move.

What the reports say

Anthropic chief Dario Amodei said he does not know exactly how Claude was used in the strike on Shajareh Tayyebeh Elementary School near Minab, Iran. He called the event “a really terrible thing to happen” and said the use case did not break Anthropic’s policy lines.

At the same time, Reports indicate Claude was integrated into the Maven Smart System, a Palantir-built targeting platform used by United States Central Command under a $1.3 billion Pentagon contract. Those reports say the system helped generate, sort, and rank targets during the Iran campaign.

Why the strike matters

The strike has become a test case for the way modern militaries use artificial intelligence. Amnesty International and Human Rights Watch reported that the strike killed at least 120 children and more than 150 people overall. Those findings have drawn international scrutiny. Those findings, along with reports that a United States-made Tomahawk missile was likely used, made the attack a global flashpoint.

Just as important, the public debate is not settled on cause. Military Times and other reports say former officials blamed stale data and human failure, not a model failure. A separate legal review also said internal reporting pointed to human error in wartime, not a problem with the deployed technology.

The accountability problem

The strongest fact in the record is also the hardest for the public to use: Anthropic says it cannot see the classified logs that would show how Claude was queried. That leaves a gap between a company that says it did not approve autonomous killing and a military system that may still have used its model inside the targeting chain.

ICYMI O/N

IRAN:

Oil and bond yields climbed into Wednesday after the U.S. pounded sites along Iran’s coast with airstrikes and blocked its ability to sell oil legally on Tuesday in a fierce response to Tehran’s recent attacks on ships near the Strait of Hormuz. Iran responded…

— trap_zack (@ZackEiseman) July 8, 2026

That gap matters because it lets each side keep its own story. Anthropic can say a human made the final call. Critics can point to a system built to speed up target selection and say increasing operational speed may also increase the risk of human mistakes if verification procedures fail. In a war setting, that kind of split can make it difficult to determine who ultimately made the decision.

What comes next

The biggest unanswered question is simple: what did Claude actually do, if anything, in the Minab strike? The available reporting does not prove that the model picked the school as a target. It does show that Claude was part of a wider military targeting system already tied to the strike campaign.

That is why calls for logs, audit records, and analyst testimony now sit at the center of the story. Without those records, the public is left with a familiar pattern: a deadly civilian loss, a powerful technology company, a military process kept behind classified walls, and no clear answer on where human choice ends and machine help begins.

Sources:

feedpress.me, washingtonpost.com, thenextweb.com, msukhareva.substack.com, forbes.com