AI systems are accelerating US military decision-making in Iran, where the Pentagon reports striking over 2,000 targets in just four days. The technology, powered by Palantir's Maven Smart System and Anthropic's Claude model, processes intelligence data from drones and satellites to generate strike options in seconds rather than hours. This marks the first battlefield use of 'frontier' generative AI models.

Military officials say the AI shift enables unprecedented operational speed, with systems now supporting the entire 'kill chain' from target identification to battle damage assessment. The technology has expanded rapidly, with Maven now used by over 20,000 military personnel across 35 entities, potentially reaching 50,000 users. NATO has also signed on to use the system in 2025.

The accelerated targeting raises serious concerns about oversight and accountability. A girls' primary school bombing in Minab highlights the risks of rapidly generated or improperly vetted targets. Critics question how to ensure meaningful human control when AI systems make millions of computations per second, particularly as the US and Israel have struck over 20,000 non-military buildings in Iran according to the country's Red Crescent.