The Pentagon's AI Is No Longer Hypothetical
The day's biggest signal: artificial intelligence has moved from the Defense Department's labs to its operations. The U.S. government acknowledged in a recent legal filing that Elon Musk's Grok chatbot, built by xAI, was used to support military strikes against Iran. The Pentagon has since confirmed it used Grok to help plan and carry out those strikes — a striking admission that a commercial AI model now sits inside the kill chain.
That disclosure lands alongside evidence of how deeply AI is already woven into the building. The Pentagon says it is using AI to draft the reports it is legally required to submit to Congress, with roughly 1.5 million users now active across the department. And OpenAI is set to bring ChatGPT to the DoD's generative-AI platform, GenAI.mil, in 'early July,' a company official said Tuesday — putting the tool within reach of some 3 million service members.
The through-line: tools that began as office productivity software and consumer chatbots are now standard-issue across the world's most powerful military.
Ukraine Becomes AI's Live-Fire Range
Nowhere is the shift more visible than over Ukraine. Kyiv has turned the skies above its cities into a testing ground, fielding AI-powered systems to detect and shoot down the Russian drones that have come to define the war. The lessons are not staying local. More than 500,000 hours of drone footage from the conflict have been folded into a military AI training dataset — half a million hours of real-world combat data now available to developers building defense applications worldwide. What Ukraine learns under fire is becoming a global resource, accelerating the next generation of autonomous and semi-autonomous weapons.
That fits a broader theme analysts are calling 'physical AI' — intelligence embedded not in chatbots but in machines that move, perceive, and act in the real world. The argument from this corner of the field is blunt: nations that ignore physical AI in warfare do so at their peril.
Washington Turns on Anthropic
In a sharp reversal, the company that asked to be regulated is now feeling the squeeze. Anthropic, maker of the Claude assistant, has spent years publicly urging the government to set rules for AI. According to CNBC, the Trump administration is responding — but not in the way Anthropic hoped. The White House has moved to restrict government access to the firm's most capable models, a decision reportedly spurred after Amazon raised concerns. The episode is a reminder that calling for oversight and welcoming the form it takes are two very different things.
The China Front: DeepSeek's Reprieve
The administration's posture toward Chinese AI is proving less than steady. Washington was preparing to add DeepSeek to its Entity List — the trade blacklist that would sharply curb the firm's access to American technology — before hitting pause. The hesitation underscores how tangled export-control policy has become as AI rivalry with Beijing intensifies.
A Helping Hand for xAI
Finally, the government stepped in on the other side of a courtroom. The Justice Department has asked a court to dismiss an NAACP lawsuit against Musk's xAI over its Colossus 2 data center. Coming the same day as the Pentagon's Grok confirmation, the move highlights how closely some AI firms' commercial and national-security fortunes are now intertwined with federal interests.
The Takeaway
A single day captured the speed of the shift: AI helping plan strikes, drafting laws' required reports, learning from half a million hours of combat, and dividing the government over which companies to trust and which to restrain. The technology is no longer arriving at the front. It is already there.