Musk’s Grok AI directed 2,000+ strikes in Iran after SpaceX merged with xAI

For the space and defense community, the revelation that Grok was used in Operation Epic Fury is not just an AI story—it’s a cold launch into a new era of integrated space-AI warfare. SpaceX’s absorption of xAI in February put the same company behind Starlink, Starship, and the targeting brain of Project Maven. The legal brief means that a Musk-controlled commercial space giant now provides the AI layer for lethal strikes, tightening the convergence of launch vehicles, satellite constellations, and tactical decision-making.

The U.S. government has disclosed that Elon Musk's Grok AI model was used in a major military operation against Iran, marking a dramatic escalation in the deployment of commercial artificial intelligence for lethal targeting. The revelation came in a June 15, 2026, Department of Justice legal brief defending xAI’s energy infrastructure against an environmental lawsuit, and it immediately raises profound questions about the militarization of AI, the erosion of ethical safeguards, and Musk’s deepening entanglement with the defense apparatus. The same brief contains sworn testimony from Pentagon AI chief Cameron Stanley, who confirmed that Grok has been integrated into Project Maven—the Pentagon’s flagship AI targeting program—and that its 'Gov Model' was instrumental in Operation Epic Fury, during which U.S. forces delivered over 2,000 munitions to 2,000 distinct targets within 96 hours.

“

SpaceX’s absorption of xAI in February put the same company behind Starlink, Starship, and the targeting brain of Project Maven.

The Grok deployment represents a sudden pivot from Anthropic’s Claude model, which had previously powered Maven. Anthropic, known for its constitutional AI safety principles, refused to allow its technology to be used in military applications, leading the U.S. government to terminate its contracts at the end of February 2026. This abrupt switch to Grok—developed by Musk’s xAI and folded into his space exploration company SpaceX only weeks earlier—has effectively circumvented the kind of voluntary ethical boundaries that had begun to define responsible AI development. The merging of xAI into SpaceX creates a uniquely integrated tech-defense giant, one that now supplies its own AI directly to a space-focused enterprise that already holds sensitive national security contracts. That convergence blurs the line between civilian commercial space ventures and offensive military operations in a way not seen before.

The lawsuit that forced this disclosure—filed by the NAACP against xAI over alleged Clean Air Act violations—centers on dozens of gas turbines powering an xAI data center in a predominantly Black neighborhood, which the civil rights group says are operated without permits. The DOJ’s brief argued that shutting down the data center would imperil national security because it supports AI innovation critical to the Department of War’s military operations. This environmental-justice-versus-national-security framing puts a stark spotlight on the hidden costs of AI infrastructure: the communities bearing the pollution burden, the lack of transparency, and the ability of defense imperatives to override environmental and public health concerns. xAI maintains the turbines are temporary and mobile, and thus exempt from regulation, a legal stance that the lawsuit must now weigh against the newly revealed combat role of Grok.

For the broader AI and defense sectors, the immediate implications are jarring. Anthropic’s principled refusal is now a case study in how quickly ethical guardrails can be cast aside when a more pliable vendor emerges. Musk, who has at times voiced caution about AI’s existential risks, has now positioned Grok as the engine of a lethal targeting system, with Stanley praising the 'greatly increased operational efficiency' of the Grok Gov Model. The efficiency mentioned is not abstract: it translates directly into a reduced time-to-strike and an expanded kill box, raising concerns about the speed of future conflicts and the degradation of human oversight in life-or-death decisions. The reference to over 2,000 distinct targets in 96 hours suggests a scale and tempo that would be difficult for traditional human-in-the-loop processes to match, underscoring the acceleration of warfare by AI.