- A Department of Justice brief defending xAI’s data centers reveals the commercial chatbot has been absorbed into “mission-critical operations” as a matter of paramount national security
By SCM National Security Correspondent
WASHINGTON — In a startling disclosure that underscores how deeply artificial intelligence has been woven into modern lethal warfare, the United States government has admitted that a military-grade version of Elon Musk’s chatbot, Grok, was used to orchestrate thousands of missile strikes against targets in Iran.
The revelation, hidden within a Department of Justice legal brief filed on June 15, marks the first official confirmation by the federal government that xAI’s commercial technology has been fully absorbed into live-fire combat operations.
According to a sworn declaration by Cameron Stanley, the Pentagon’s chief digital and artificial intelligence officer, the software—dubbed the “Grok Gov Model”—was integrated directly into Project Maven, the military’s flagship AI targeting program.
Working in tandem with Maven Smart Systems, a data analysis platform co-developed with Palantir, the AI system “enabled U.S. forces to deploy over 2,000 munitions to 2,000 distinct targets within 96 hours during Operation Epic Fury,” Mr. Stanley stated.
The admission came not via a planned press briefing, but through a routine federal court case in Mississippi. The Department of Justice submitted Mr. Stanley’s testimony to fend off an environmental lawsuit filed by the NAACP, which seeks to halt unpermitted, gas-fired turbines powering xAI’s massive “Colossus” supercomputer data center.
Federal prosecutors argued that any disruption to the power grid fueling Mr. Musk’s data center would directly threaten American defense capabilities, rendering the continued, unhindered operation of Grok “a matter of paramount national security.”
In his testimony, Mr. Stanley praised the technology for delivering “greatly increased operational efficiency,” affirming that the chatbot is now “equipped to support mission-critical operations.”
He went further, comparing the necessity of computing infrastructure to world-war era industrial manufacturing. “The critical ability to field its data facilities at massive scale is as foundational to our modern defense posture as traditional munitions production,” Mr. Stanley asserted.
The Pentagon’s abrupt transition to Grok follows a turbulent chapter in the military’s relationship with Silicon Valley. Earlier this year, the Department of War abruptly terminated its existing contracts with Anthropic, a rival AI firm.
Anthropic executives had reportedly refused to allow their marquee model, Claude, to be used for fully automated lethal strikes or the mass surveillance of citizens.
Following Anthropic’s exit, Defense Secretary Pete Hegseth signaled a cultural and operational shift in how the military selects its tech partners, explicitly stating that future military AI systems would operate “without ideological constraints” and that the Pentagon’s AI “will not be woke.”
Mr. Musk, a close ally of President Trump whose company xAI was recently folded into SpaceX, designed Grok precisely as a counterweight to what he termed “woke AI.” However, the sudden deployment of the software in highly sensitive conflict zones has raised intense ethical and operational alarms.
The speed of the AI-driven campaign—hitting 2,000 separate targets in just four days—has drawn sharp criticism from humanitarian groups.
A concurrent report by Yahoo News suggested that U.S. military investigators believe AI-driven targeting workflows were linked to an airstrike on a girls’ school in Minab, Iran, which resulted in the deaths of at least 175 civilians, mostly children.
Legal and tech experts warn that outsourcing lethal targeting decisions to generative large language models opens an unpredictable frontier in warfare. Unlike traditional deterministic software, generative AI can “hallucinate” or misinterpret unstructured data, making accountability difficult to pin down when civilian casualties occur.
Within tech circles, the military embrace of Grok has sparked internal revolts reminiscent of the 2018 employee protests over Google’s early involvement in Project Maven. At Google, more than 600 employees recently signed petitions demanding the company halt provisioning AI models for classified military networks.
As the legal battle over xAI’s Mississippi power plant continues, the Pentagon has made its stance clear: the future of American warfighting is now tied inextricably to commercial AI, and the algorithms guiding American missiles are no longer just tools, but core elements of the nation’s defense.
For a deeper dive into how artificial intelligence is changing modern warfare, you can watch this detailed report on AI in the Iran war which outlines the ethical dilemmas and corporate friction behind these military advancements.
