The Trump administration relied on a special version of Elon Musk’s Grok chatbot to help launch thousands of missiles in Iran, the U.S. Defense Department said Wednesday in a formal statement. The disclosure came in sworn testimony by Cameron Stanley, the Pentagon’s chief digital and AI officer, in a Mississippi court case tied to a lawsuit against Musk and his company xAI over its data center operations in the state.
Stanley said keeping Grok running is a matter of “highest national security,” and that it helped direct more than 2,000 munitions at thousands of targets within 96 hours. He also said the Mississippi data center and others are “well positioned” to provide a “critical jump” in energy capacity during conflict or other urgent national-security situations. The Pentagon said Grok is one of four AI models currently able to support national-security uses and one of three products ready for highly classified critical missions.
The testimony marked the first explicit admission by a senior U.S. official that the military used an AI chatbot in bombing operations in Iran. U.S. media also revisited early-war reports about a strike on an Iranian girls’ school in Minab that killed at least 175 people, the deadliest civilian incident in the war. Analysts said the attack likely resulted from AI-assisted targeting combined with human error, including a failure to check whether maps had been updated.
The Pentagon has also been using the Maven Smart system, run by the National Geospatial-Intelligence Agency and originally powered by Anthropic’s Claude model before that relationship deteriorated. Court filings say the military is now using a special Grok package called Grok Gov, built for federal agencies with capabilities not found in other front-end AI models. Meanwhile, Democratic senators are preparing legislation to curb military AI use, including a bill by Kirsten Gillibrand that would keep life-or-death decisions in human hands and ban AI use for nuclear weapons, civilian surveillance, and autonomous weapons. The Pentagon is separately locked in a legal fight with Anthropic after failing to agree on usage limits, and has labeled the company a national-security supply-chain risk.
