What happens when artificial intelligence is given direct influence over where missiles strike? In Iran, we are now finding out in real time. The Trump administration used Elon Musk’s Grok artificial intelligence to guide thousands of strikes, and for the first time, a senior Defense Department official has admitted it in black and white.
The admission did not come at a press conference or in an annual report, but emerged from a declaration in a court case. In it, the Pentagon is defending Musk and his company xAI, which is being sued by the NAACP over air pollution in predominantly Black communities. And it was precisely to protect Musk’s data center that the Pentagon disclosed information it would otherwise most likely have kept under wraps.
What exactly the court filing says
The declaration was signed by Cameron Stanley, the Defense Department’s chief official for digital technology and artificial intelligence. He wrote that Grok’s continued operation is “a matter of the highest national security.” He also provided specific, chilling figures. According to him, Grok “enabled U.S. forces to deploy more than 2,000 missiles and similar munitions against 2,000 different targets within 96 hours” during an operation called Epic Fury. Stanley described this as evidence of significantly greater combat effectiveness delivered by the so-called Grok Gov Model, a version of the chatbot tailored to federal agencies.
Grok is one of four artificial intelligence models that, according to Stanley, are “currently capable of supporting national security applications.” It is also among three products equipped to support “mission-critical operations” in a top-secret environment. Stanley further claims that the Grok Gov Model offers capabilities that “cannot be found in any other leading artificial intelligence model.”
This appears to be the very first open confirmation from an administration official that the government is using Musk’s artificial intelligence to bomb Iran.
Dead children and suspicions that both technology and people failed
It is important to be precise here. Targets for Operation Epic Fury were not identified by Grok alone. The National Geospatial-Intelligence Agency’s Maven Smart System also assisted, using artificial intelligence to compile data into a clear dashboard and support officials in their decision-making. These products do not generate targets on their own. They operate within Maven and search for potential points of interest for military intelligence. A human therefore still has the final say. At least for now.
And now for the most difficult part. The use of artificial intelligence on the battlefield came under scrutiny after U.S.-led attacks killed hundreds of civilians, including children.
U.S. military investigators believe that American forces were most likely responsible for the strike on a girls’ school in the city of Minab. At least 175 people were killed, most of them children. Analysts and human rights advocates consider it the deadliest incident involving civilian casualties since U.S. and Israeli forces began attacks on the ground in February.
What caused the tragedy? Some experts warn that artificial intelligence-assisted targeting may have played a role. This was compounded by human error, as no one verified whether the target maps were up to date. A deadly combination.
Dispute over a data center that “breathes” into a Black neighborhood
Let us return to why the Pentagon disclosed this sensitive data in the first place. It was trying to save a data center near Memphis.
The Trump administration is asking a federal court in Mississippi to throw out the NAACP’s lawsuit. The organization claims that xAI is violating the Clean Air Act by operating dozens of gas turbines without the required permits. Specifically, at least 57 turbines are powering the Colossus 2 data center without the emissions controls required by law.
The Justice Department wants the case dismissed, arguing that access to the Colossus 2 data center is “essential” to national security. Stanley described it as vital for targeting, intelligence, emergency response, and recruitment tasks. Data centers can reportedly provide a “critical surge” in power capacity in the event of “armed conflict or other extraordinary circumstances.” According to the Justice Department, the Pentagon would be “seriously” harmed if the court prevented xAI from continuing to be deployed, fine-tuned, and improved across government agencies.
Politicians want an emergency brake
Some Democrats in Congress are now pushing to restrict the military’s deployment of artificial intelligence. They are responding to senior officials’ refusal to investigate civilian deaths that stricter oversight of artificial intelligence might have prevented. Senator Kirsten Gillibrand has introduced a bill that would ensure human commanders continue to make life-and-death decisions. It would completely prohibit the use of artificial intelligence for nuclear weapons, domestic surveillance, and autonomous weapons systems.
“The most consequential decisions affecting our national security and the lives of our service members must always be made by people, not machines with no accountability whatsoever,” she said. She warned that the Pentagon is moving toward deploying extraordinarily powerful technology without reasonable safeguards, which could have catastrophic consequences. “We must act now. Not to slow technological progress, but to establish clear rules that keep people at the helm and make the use of artificial intelligence in warfare smart and safe.”
Sources: theinformation.com, independent.co.uk and thehill.com
