Grok assisted in targeting 2,000 locations in Iran. Now, the resulting pollution is considered a matter of 'national security.'

      The Pentagon states that Grok aided in striking 2,000 targets in 96 hours, claiming that the polluting power plant associated with it is of utmost national security importance. The troubling aspect lies in the interconnection of these two assertions.

      This revelation did not emerge from a press release or Pentagon briefing; instead, it was included in a statement submitted to a federal court in Mississippi regarding air pollution. There, while defending Elon Musk’s xAI against a lawsuit related to the Clean Air Act, the Pentagon’s chief digital and artificial intelligence officer made a statement that is shocking for anyone who reads it.

      He noted that the chatbot known as Grok had facilitated the firing of over 2,000 munitions at 2,000 separate targets in Iran within a 96-hour timeframe and that its ongoing operation was essential for national security.

      Take a moment to reconsider that, as the context makes it all the more significant. A government official revealed that a consumer AI tool was utilized to conduct bombings in a foreign country, not to inform the public but to ensure the functioning of a data center. In the government’s view, the issues of targeting and environmental impact are linked: Grok is vital for the war, the war matters for the nation, so the power plant that sustains Grok cannot be shut down, regardless of legal permit requirements.

      This moment merges two separate concerns regarding military AI. The first is the fact that a chatbot designed for social media is now integrated into a system that causes harm. The second is that the owner of this chatbot plays a role in the government decisions regarding its deployment. Each concern is serious on its own.

      Together, they paint a picture of a system where the usual safeguards—legal, environmental, ethical—are compromised in favor of one individual and one company’s convenience.

      To begin with the chatbot, Grok is allegedly one of only four AI models deemed capable of supporting national security applications by the Pentagon and one of three approved for critical missions in classified environments.

      It is part of the National Geospatial-Intelligence Agency’s Maven Smart System, which utilizes AI to present intelligence data to assist officials in making strike decisions. Official narratives convey a sense of reassurance: the AI does not generate targets but rather identifies points of interest for human analysts to evaluate, assuring that humans remain involved in the process.

      However, the loop did not prevent tragedy in Minab. On February 28, during the initial round of strikes, a Tomahawk cruise missile struck the Shajarah Tayyebeh girls' school in the Iranian town, resulting in numerous fatalities, including over a hundred children, making this event one of the deadliest civilian incidents of the conflict, as noted by observers.

      A preliminary military investigation revealed that U.S. forces were likely responsible for the strike and that it relied on outdated intelligence, with target coordinates based on stale data that were not cross-checked against current maps. The school had been an active educational facility throughout this period.

      This highlights the issues with employing AI in targeting decisions, which extends beyond who pulls the trigger. The risk is subtler and more challenging to regulate. An AI dashboard that ranks points of interest and assigns weapons to targets doesn’t eliminate human judgment; rather, it obscures it.

      Analysts face a sleek interface, a clear output, a target already assigned a numerical designation, which fosters a natural inclination to trust it. Automation bias is not an error in these systems; it is an expected outcome of designing tools to appear authoritative. When the foundational data is flawed, the interface renders the mistake as rigor until the missile strikes.

      Now, consider the second concern made impossible to overlook by the Mississippi filing. The lawsuit was initiated by the NAACP, alleging that xAI is operating numerous gas-fired turbines at its data centers without the necessary permits mandated by the Clean Air Act. This case focuses on a facility in Southaven, Mississippi, where the NAACP identifies 27 unpermitted turbines located near homes, schools, and churches in a predominantly Black community.

      Memphis, home to xAI’s Colossus supercomputer, ranked second in the nation for asthma-related emergency room visits in 2024. These communities are inhaling the emissions produced by the computing facilities.

      The government’s response to the lawsuit was not to argue for existing (but non-existent) permits, but to assert that the pollution is essential for military operations. The Department of Justice intervened, requesting a judge to dismiss the case, claiming that halting the turbines would "severely" hinder the Pentagon's operations.

      According to the official, the data centers are crucial for providing a significant energy capacity boost in times of armed conflict. In essence, national security has become a catch-all justification: it overrides environmental laws and the concerns of families suffering from pollution, serving the infrastructure of a private corporation without the requisite permits.

      This situation highlights the tangible conflict of interest. Musk is not an external vendor dist