Pentagon’s Chief AI Officer reveals under oath that the Grok Gov Model enabled deploying over 2,000 munitions on 2,000 targets in 96 hours during Operation Epic Fury. Anthropic had refused to participate.
On June 15, 2026, the U.S. Department of Justice filed a legal memorandum that shook the tech world: Grok, Elon Musk’s AI assistant (xAI), is officially being used by the U.S. military for targeting strikes in Iran.
The revelation came through an unexpected legal proceeding: the NAACP is suing xAI for violating the Clean Air Act over gas turbines powering a data center in Mississippi. To defend xAI, the DOJ produced a bombshell testimony from the Pentagon’s Chief AI Officer.
---
Key revelations
| Information | Detail |
|---|---|
| Model used | Grok Gov Model (government version) |
| Program | Project Maven / Maven Smart Systems (MSS) |
| Operation | Epic Fury — strikes in Iran |
| Result | 2,000 munitions → 2,000 targets in 96 h |
| Revealed by | Cameron Stanley, Chief Digital & AI Officer, Dept of War |
| Context | DOJ memo to defend xAI in environmental lawsuit |
“Grok provides essential support to the military operations of the Department of War.” — DOJ memo, June 15, 2026
---
How we got here
The story begins with Anthropic and Claude. Initially, Project Maven — the Pentagon’s AI targeting program — used Claude as its primary model.
In late February 2026, the U.S. government terminated its contracts with Anthropic. The reason? Anthropic refused to authorize the use of its tools for:
• Fully automated strikes (without human validation)
• Mass surveillance of American citizens
The Pentagon then turned to competitors: Google, OpenAI, and xAI.
“600 Google employees demanded the company not supply AI to the military for classified operations.” — International press
---
The lawsuit that triggered everything
The NAACP (National Association for the Advancement of Colored People) filed a lawsuit against xAI for violating the Clean Air Act. The subject of the dispute: the gas turbines powering xAI’s Colossus 2 data center in Mississippi.
The plaintiffs assert that these turbines, located in predominantly Black neighborhoods, emit dangerous pollutants without proper environmental permits.
“The gas turbines at the Colossus Plant power xAI’s Colossus 2 data center, which in turn powers the ‘Grok’ chatbot.” — NAACP complaint
xAI argues the turbines are temporary and mobile, therefore not subject to regulation. But the DOJ went further by invoking national security:
“The NAACP’s attempt to cut the power supporting Grok also threatens national security because… Grok provides critical support to the military operations of the Department of War.” — DOJ memo
---
The testimony that changes everything
Cameron Stanley, Chief Digital and AI Officer of the Department of War, stated under oath:
• The Grok Gov Model was integrated into Maven Smart Systems (MSS)
• During Operation Epic Fury in Iran, MSS enabled “deploying over 2,000 munitions on 2,000 distinct targets in 96 hours”
• The model possesses unique features that no other AI model offers
• Operational efficiency was significantly enhanced thanks to Grok
---
Ethical and geopolitical implications
| Question | Situation |
|---|---|
| Human validation | The DOJ did not specify whether each strike was validated by a human |
| Mass surveillance | Anthropic had refused → xAI accepted |
| Conflict of interest | Elon Musk supplies military AI via xAI, and SpaceX launches military satellites |
| Legal precedent | The “national security” argument can be used to circumvent environmental laws |
---
Where each player stands
Anthropic refused to participate and lost the contract. But in March 2026, the government had to admit that Claude was still being used for the war in Iran — a sign the transition isn’t complete.
Google faces massive internal protests (600+ employees), but continues working with the Pentagon.
OpenAI — no specific information on its participation.
xAI / Elon Musk accepted the conditions Anthropic refused. And since February 2026, xAI has merged with SpaceX, significantly strengthening the ties between AI and defense.
---
Izri’s Verdict
This revelation raises dizzying questions:
1. Military AI is no longer a hypothesis — it’s an operational reality, documented under oath.
2. The line between civilian and military AI is blurring — the same Grok model you use for coding powers strikes in Iran.
3. Ethical choices by AI companies have real consequences — Anthropic lost a lucrative contract by refusing, xAI won it by accepting.
4. The “national security defense” is a dangerous precedent — invoking national security to bypass environmental laws could create jurisprudence with unpredictable consequences.
What this means for us in Morocco: the AI race is no longer just economic or technological — it has become military and geopolitical. The AI models we use daily are the same ones deciding strike targets. The question is no longer “will AI replace our jobs?” but “who controls the AI that makes life-or-death decisions?”
Sources: The Information, Ars Technica, AFP, Straits Times, RNZ, Economic Times, DOJ court filings
