Elon Musk's AI tool Grok has been used by the U.S. military in strikes against Iran, according to a legal briefing from the American government that surfaced in recent days. The filing, submitted in a lawsuit over xAI's data center, says the Pentagon uses a dedicated version of the model for intelligence analysis and target identification.
The disclosure came in the Justice Department's defense of gas turbines operating at xAI's huge data center, which is the focus of a lawsuit by the NAACP. The civil rights group says xAI is running dozens of turbines without permits and polluting predominantly Black neighborhoods. xAI argues the turbines are temporary and mobile, so normal regulations do not apply. The government says shutting off power to the site would threaten U.S. national, economic and energy security because the facility directly supports Department of War operations.
Federal lawyers backed that claim with sworn testimony from the Pentagon's chief AI official, who said the government version of Grok is integrated into Project Maven, the U.S. military's computerized target-recognition program. Maven previously relied on Anthropic's Claude, but that contract was ended after Anthropic refused to allow use for fully autonomous strikes or mass surveillance. The Pentagon then turned to competitors including Google, OpenAI and xAI, and in March it admitted Claude was still being used in fighting in Iran during the transition.
The new system has already delivered battlefield results. The released data says the combined systems allowed U.S. forces to fire more than 2,000 munitions at 2,000 different targets in just 96 hours during one military operation. Pentagon officials praised the resulting operational efficiency. The article also notes that militaries worldwide, including the IDF, are increasingly using AI tools, while such adoption has drawn internal and public opposition, including a petition signed by more than 600 Google employees.
