Anthropic CEO Admits Claude's Iran Strike Role Is Unknown
TL;DR
- Anthropic CEO Dario Amodei said he doesn't know exactly how Claude was used in the February 28 Minab school strike.
- Claude is embedded in Maven Smart System, a Palantir-built targeting platform under a $1.3 billion Pentagon contract.
- Amodei maintained the strike didn't violate Anthropic's policies, citing human decision-making principles he admitted he cannot verify.
Anthropic CEO Dario Amodei sat for an interview on Bloomberg's The Circuit with Emily Chang and was asked about a February 28 missile strike on Shajareh Tayyebeh Elementary School in Minab, Iran, which killed an estimated 120 children. His answer was candid in a way that raised more questions than it settled: "We don't have access to, we don't know exactly how these models were used." He described the strike as "a really terrible thing to happen" and maintained the use case did not violate Anthropic's policies.
The trouble is the argument that follows. Amodei also said the principle the company has established — that "a human makes the final decision" — was "obeyed" in the Minab strike. Saying both things at once is difficult to sustain. If you do not know how the model was used, you cannot know whether the human-in-the-loop principle held. Hamza Chaudhry of the Future of Life Institute called that requirement a "rubber stamp." Jack Shanahan, the retired Air Force lieutenant general who created Project Maven, put the targeting problem more bluntly: "May have a thousand targets, but are they the right targets?"
Claude is embedded in the Maven Smart System, an AI-assisted targeting platform built by Palantir under a $1.3 billion Pentagon contract. The system reportedly struck 1,000 targets in the first 24 hours of operations and reached 13,000 targets by April 6. Defense Secretary Pete Hegseth promised a "thorough probe" in March, but the outcome of that investigation was not addressed in the Bloomberg interview.
What the reporting does not give you is whether Anthropic has any contractual right to observe how Palantir deploys Claude in Maven, or what would actually constitute a policy violation in a system the company cannot see. That gap is not unique to Anthropic — it is the structural problem facing any AI developer that licenses a foundation model to a defense contractor and has no telemetry on downstream use. If contract terms come under congressional scrutiny, or the investigation produces public findings, those disclosures will matter well beyond this incident.
Shared on Bluesky by 2 AI experts
-
Let’s check in on the “ethical” ai company….
View on Bluesky →
Originally reported by bloomberg.com
Read the original article →Original headline: Anthropic CEO Doesn’t Know If Claude Used in Iran School Strike