The first time the U.S. government admitted using Elon Musk's Grok to help bomb Iran, it did so in a court filing about air pollution.

Cameron Stanley, the Pentagon's chief digital and artificial intelligence officer, swore in a Mississippi court that Grok's continued operation is "a matter of paramount national security." The statement was filed to defend xAI against an NAACP lawsuit alleging its data center runs dozens of gas turbines without Clean Air Act permits, polluting majority-Black neighborhoods. The Justice Department's argument was blunt: shutting off the turbines "threatens American national, economic and energy security" because they power AI used in military operations. [1]

So how did missiles end up in a pollution case? To prove the data center mattered, Stanley described what the software it powers actually does.

Here is the part the viral headlines get wrong. The widely shared claim is that "Grok fired 2,000 missiles at Iran." What Stanley actually said is that the Maven Smart System "enabled US forces to deploy over 2,000 munitions to 2,000 distinct targets within 96 hours" during an operation called Epic Fury. [1] Grok is one of four AI models cleared to support national-security work, and one of three approved for top-secret operations. [2] It is a component inside Maven, not the thing pulling the trigger. Stanley credited "the greatly increased operational efficiency made possible by the Grok Gov Model." That is efficiency language, not an admission that a chatbot launched 2,000 strikes.

The confirmed core is still remarkable: a commercial chatbot is now embedded in the live targeting pipeline for U.S. strikes on Iran. Maven was originally powered by Anthropic's Claude, until the government terminated that contract in February after Anthropic refused to allow fully automated strikes or mass surveillance. [1] The Pentagon then turned to Google, OpenAI and xAI.

The scrutiny isn't abstract. Investigators believe American forces likely struck a girls' school in Minab that killed at least 175 people, mostly children, the deadliest civilian incident of the war, with analysts pointing to AI-driven targeting and a failure to verify whether target maps were current. [2]

Senator Kirsten Gillibrand is pushing a bill to keep humans in charge and bar AI entirely from nuclear weapons, domestic surveillance and autonomous weapons. "The most critical decisions affecting our national security and the lives of our service members must always be made by human beings, not unaccountable machines," she said. [3]

The Pentagon has not separately confirmed or denied the operational details beyond the filing itself, which is the one thing observers are now watching for. [4]

Sources

Primary source: Pentagon used Elon Musk’s Grok AI to fire 2,000 missiles at Iran, official says - The Independent

  1. Musk’s AI, Grok, was used in strikes in Iran, reveals Pentagon - Le Monde.fr — reporting
  2. Pentagon used Musk’s Grok AI to fire thousands of missiles at Iran, court filing says - Ynetnews — reporting
  3. Pentagon officials say Grok AI was used to fire 2,000 missiles at Iran - The American Bazaar — reporting
  4. Pentagon used Musk’s Grok AI to launch 2,000 missiles at Iran: court filing - Crypto Briefing — reporting