The Pentagon is now on record: a commercial chatbot helped bomb thousands of targets in Iran. The admission came in a sworn filing by Cameron Stanley, the Defense Department’s top AI official. The chatbot was Grok, built by Elon Musk’s xAI.
The strikes happened over a short period. The system was efficient.
This is not a theoretical debate anymore. The technology is live. It is killing people.
The filing lands in the middle of a separate legal fight. The NAACP is suing xAI, arguing its data centers are polluting Black communities in Mississippi and Tennessee.
The federal government wants that lawsuit thrown out. The administration’s argument: these facilities are vital national-security infrastructure. They provide emergency power.
They preserve a technological edge. That calculus is now laid bare. Operational efficiency against environmental and social cost.
The Pentagon chose the first. Grok Gov, the specialized version used in the strikes, is one of a handful of AI tools cleared for top-secret work.
The Pentagon says it can support sensitive, mission-critical operations. In Iran, it did. Democratic Sen. Kirsten Gillibrand has been pushing to slow this down.
She has warned about deploying powerful AI without adequate guardrails. Her concerns now have a real-world case to point to.
The stakes are concrete. The U.S. military believes AI gives it a critical advantage in future conflicts. That advantage, they argue, depends on using these tools now, in real operations, against real targets.
But the same speed and efficiency that makes Grok attractive for strikes also makes it harder to control. A system that can process thousands of targets quickly does not pause for moral questions. It does not second-guess.
There is no public record of what checks were in place. No independent audit.
No congressional oversight of the specific strikes. The filing from Stanley is a disclosure, not a defense. It states what happened.
It does not justify it. The NAACP case forces a different question.
If the data centers that power these AI tools are harming nearby communities, the cost of the technological edge is not just paid in foreign airstrikes. It is paid in polluted air and water in Mississippi and Tennessee. The administration argues those harms are secondary to national security.
That argument is now before a federal judge. Grok Gov is not a hypothetical. It is not a pilot program.
It is a deployed weapons system. The Pentagon has said as much in court.
The strikes in Iran were not a test. They were operations. Hundreds of targets.
A short window. A chatbot making it possible.
The broader military adoption of AI is accelerating. The Pentagon is not hiding that. They are filing it in court documents.
The question now is not whether the technology works. It clearly does. The question is what happens when it does not.
What happens when the system makes a mistake at that speed, at that scale. What happens when the data center powering the AI poisons a neighborhood.
What happens when Congress tries to catch up. Gillibrand is trying. The NAACP is trying.
The Pentagon is arguing both should step aside. The judge will decide the lawsuit.
The strikes in Iran are already over. The technology is still running.
