What Happened
On March 6, 2026, Futurism reported that the Pentagon refused to answer whether AI was used in the targeting decision that led to the bombing of Shajareh Tayyebeh girls' school in Minab, Iran. The attack killed 165 elementary students and staff, aged 7 to 12. The school was hit twice - the second strike targeted first responders and parents who had arrived to collect their children.
The question is not hypothetical. The Wall Street Journal previously reported that the Pentagon has been using Anthropic's Claude AI model to assist in planning military strikes on Iran. When Futurism contacted the Pentagon directly about AI involvement in selecting the school as a target, they were referred to US Central Command, which responded: "We have nothing for you on this at this time."
The strikes are part of a broader campaign that has claimed over 1,000 lives in under one week.
Why It Matters
This story forces a question that the AI industry has been able to sidestep in commercial contexts: what happens when the tools built for productivity and knowledge work are applied to lethal targeting decisions?
The comparison to Israel's "Lavender" AI system, reported in April 2024, is direct. That system identified 37,000 Palestinians as potential targets in Gaza, with one Israeli military source noting that human operators spent roughly 20 seconds per target decision - functioning primarily as "a stamp of approval" rather than a genuine check on the AI's output.
The Pentagon's non-answer is itself significant. A flat denial would be simple if AI was not involved. The referral to CENTCOM, which then offered nothing, follows a pattern of institutional deflection on AI's role in lethal operations.
For anyone working in or around AI, this is no longer an abstract ethics debate. Commercial AI models built by companies that market themselves on safety and responsible use are reportedly being integrated into military kill chains.
Our Take
This is the hardest story to write on an AI productivity site, but ignoring it would be dishonest. The same Claude model that helps people write emails, analyze documents, and build software is reportedly being used to plan airstrikes.
Anthropic has built its brand on AI safety. Their Responsible Scaling Policy and Constitutional AI research have positioned them as the cautious alternative in the AI race. That positioning becomes difficult to maintain when your flagship model is linked - even indirectly - to targeting decisions that kill children.
The "we have nothing for you" response from CENTCOM is not reassuring. It leaves open every worst-case interpretation. If AI was not involved in this specific targeting decision, saying so would cost nothing. The silence suggests the answer is either yes, or close enough to yes that confirming it would create problems.
For users who choose AI tools partly based on the ethics of the companies behind them, this is relevant information. It does not change what Claude can do as a product. But it does change the context in which you are using it.
The broader issue is that there is currently no public framework governing how commercial AI models get repurposed for military targeting. Until that changes, every major AI company's safety commitments exist in a space where their technology can be quietly redirected toward lethal applications with no transparency and no accountability.