Claude is the US military's core AI tool for operations in Iraq, striking over a thousand targets on the first day through the Maven system.

According to 1M AI News monitoring, Anthropic's Claude is the core AI tool used by the US military in the "Operation Epic Fury" against Iran, referred to as the "most advanced artificial intelligence" ever used by the US military in warfare. Claude is integrated into the Maven Smart System developed by Palantir, accessing 179 data sources including satellite imagery, surveillance systems, and classified intelligence to generate strike targets, precise coordinates, prioritize targets, and assess strike effects, supporting strikes on over 1000 targets on the first day of the operation. As of May 2025, over 20,000 military personnel have used this system.

Claude's battlefield deployment comes at a time of strained relations between Anthropic and the Pentagon. Prior to the start of the operation, Trump signed an executive order hours earlier demanding federal agencies to cease the use of Anthropic products within six months, yet the military continues to use them. Pentagon officials stated that if Amodei attempts to cut off access, government powers will be invoked to retain this technology: "Regardless of his moral stance, we will not let his decision take away even a single American life."

Hegseth confirmed on Wednesday during a Pentagon press briefing that AI autonomous systems were deployed in the operation: "We have a significant number of autonomous systems, including drones equipped with integrated intelligent AI, many of which I cannot detail here." Chairman of the Joint Chiefs of Staff Dan Caine stated that over 2000 targets have been struck so far, with an 86% decrease in Iranian ballistic missile launches since the beginning of the conflict. Researcher Paul Scharre from the New American Security Center stated that the key paradigm shift is that AI enables the US military to generate strike plans at "machine speed rather than human speed," while also cautioning that "AI can make mistakes, and we need humans to review the outputs."