Not a leak. Not a rumor from a defense podcast. A sworn court declaration from Cameron Stanley, the Pentagon’s chief digital and artificial intelligence officer — filed in a Mississippi environmental lawsuit, not a defense briefing — confirms that xAI‘s Grok Gov Model was integrated into U.S. targeting systems during Operation Epic Fury in Iran. According to that filing, Grok helped American forces deploy more than 2,000 munitions against 2,000 distinct targets in 96 hours, as reported by The Hill. This is the clearest public admission yet that a named commercial AI model played a direct role in an active, lethal military campaign — raising urgent questions about hypersonic missiles and next-generation strike systems already reshaping modern warfare.

What Grok Actually Did (and Didn’t Do)

The distinction between “decision support” and “de facto targeting authority” is narrower than the Pentagon would like you to believe.

Grok didn’t press a button. But calling it a passive bystander stretches credulity. Embedded inside Project Maven’s Smart System, Grok Gov processed classified and open-source intelligence, flagged potential targets, and compressed planning timelines that traditionally took days into hours. Pentagon doctrine still insists humans remain “in the loop” for lethal decisions. Critics counter that when AI accelerates the path from data to strike execution this aggressively, the practical window for human deliberation narrows to something barely recognizable.

Stanley’s declaration describes Grok Gov as one of four AI models capable of national security applications and one of three rated fit for mission-critical operations in top-secret environments. The Pentagon’s GenAI.mil initiative aims to deploy Grok and similar models to roughly three million military and civilian users. Meanwhile, U.S. military investigators believe American forces were likely responsible for a strike on a girls’ school in Minab that killed at least 175 people, mostly children, according to The Hill. AI-assisted targeting workflows are under scrutiny, though direct causation to Grok has not been established.

175 people. Mostly children. That number belongs in the headline, not a subordinate clause. Remember Google employees walking out over Project Maven in 2018? That moral reckoning feels distant now.

“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.” — Sen. Kirsten Gillibrand

The Data Center Nobody Wants to Talk About

The same filing shielding Grok from legal disruption is also the government’s defense of a pollution lawsuit targeting a Black community in Mississippi.

The lawsuit context sharpens the contradiction. Stanley’s declaration landed in a federal case where the NAACP alleges xAI’s Colossus 2 data center in Mississippi runs 57 gas turbines without Clean Air Act permits, disproportionately burdening Black communities. The DOJ argues disrupting xAI’s operations would “severely” impact national security. The AI arms race carries a pollution surcharge borne by communities that had no voice in the arrangement.

The contrast with Anthropic is instructive. When Anthropic refused Pentagon terms that left open domestic surveillance or autonomous drone use, the Pentagon designated it a “supply-chain risk to national security.” xAI said yes. Now Grok sits in the kill chain, and the government fights pollution lawsuits on its behalf. The use of a surveillance app to track and target individuals illustrates exactly the kind of unchecked AI deployment that Gillibrand’s legislation seeks to prevent.

Grok Gov provides capabilities “found in no other frontier AI model” and is “vital to maintaining our technological advantage against adversaries.” — Cameron Stanley, Pentagon CDAO, sworn declaration

If 2,000-target AI-assisted campaigns become standard procedure, the tempo of future conflicts accelerates — and so does the margin for catastrophic error. Senator Gillibrand’s proposed legislation would mandate human control over life-and-death decisions and ban AI from nuclear command, domestic surveillance, and fully autonomous weapons systems. Whether that bill becomes the floor of accountability or the ceiling depends entirely on whether Congress treats this moment as an alarm or background noise.