Sam Altman is the kind of CEO who wants to be both the conscience of technology and the defense contractor of the future — and somehow thinks nobody will notice when those two characters show up in the same week wearing different outfits.
Monday: principled civil libertarian. Tuesday: Pentagon subcontractor. By Wednesday, the internet had already stitched together the contradiction and slapped a community note on it like a toe tag.
The Monday Performance
Altman took to social media Monday with the kind of post designed to make privacy advocates swoon. He announced that OpenAI was amending its contract with the federal government to explicitly prohibit domestic surveillance of U.S. citizens.
The language sounded like it was drafted by the ACLU’s PR department: “Consistent with applicable laws, including the Fourth Amendment to the United States Constitution, the AI system shall not be intentionally used for domestic surveillance of U.S. persons and nationals.”
He added that the Department of War agreed the NSA and other intelligence agencies would not use OpenAI services. He talked about “democratic process” and “civil liberties.” He name-checked “alignment, democratization, empowerment, and individual agency” as his guiding principles.
It was a beautiful performance. The kind of Silicon Valley virtue signal that gets 50,000 likes and a glowing writeup in Wired.
It lasted about eighteen hours.
The Tuesday Reality
The next day, in an internal meeting with staff, Altman delivered a very different message. CNBC reported that he told employees, point-blank, that OpenAI does not “get to make operational decisions” about how the military uses its AI.
“So maybe you think the Iran strike was good and the Venezuela invasion was bad,” Altman reportedly said. “You don’t get to weigh in on that.”
The Pentagon uses OpenAI’s tools. Secretary of War Pete Hegseth makes the final calls. OpenAI builds the product. The government decides what to do with it. End of discussion.
That’s not a nuanced evolution of position over time. That’s a Monday press release and a Tuesday reality check delivered to the same company within the same news cycle.
The Community Note Heard ‘Round the Internet
X users caught the whiplash immediately. Altman’s Monday post got hit with a community note pointing out that his public statements were “the opposite” of what he told employees the next day.
And it’s hard to argue with the characterization. Monday: the NSA can’t touch our product, we believe in the Fourth Amendment, civil liberties are sacred. Tuesday: you don’t get a vote on military operations, the Pentagon makes the calls, and I hope they keep working with us even when our “safety stack annoys them.”
That last phrase — “even if our safety stack annoys them” — is the most revealing eight words Altman has ever spoken. It tells you that the “safety” language and the Fourth Amendment posturing aren’t principles. They’re friction. Obstacles he hopes the Pentagon will tolerate in exchange for access to the most powerful AI tools on the planet.
He’s not protecting civil liberties. He’s managing the optics of not protecting them.
The Real Calculation
Here’s what’s actually happening beneath the press releases and the principles. OpenAI has a massive government contract. The Department of War is one of the biggest potential customers for AI technology in the world. Military applications of large language models — intelligence analysis, logistics, communications, decision support — represent a revenue stream that dwarfs anything the consumer market offers.
Altman wants that contract. He needs that contract. OpenAI’s burn rate is legendary, and government money is the most reliable money on earth.
But he also wants to be the good guy. He wants the San Francisco tech press to write about him as a thoughtful leader grappling with the ethics of AI in warfare. He wants his employees — many of whom came from academic and progressive backgrounds — to feel comfortable building tools that might end up guiding military strikes.
So he gives them Monday. The Fourth Amendment. The NSA prohibition. The principled stand.
And then he gives the Pentagon Tuesday. You don’t get a vote. The Secretary decides. We hope they keep working with us.
It’s a two-audience strategy, and it works right up until both audiences read the same internet.
The Bigger Problem
This isn’t just about Altman’s hypocrisy, although that’s rich enough to fill a semester of business ethics case studies. It’s about the fundamental tension at the heart of AI development.
These companies are building the most powerful information tools in human history. The same technology that writes your kid’s homework can analyze intelligence intercepts, process satellite imagery, and support targeting decisions in a military theater. The line between consumer AI and military AI isn’t a line — it’s a suggestion.
Altman wants to draw that line on Monday and erase it on Tuesday. He wants his employees to believe the Fourth Amendment clause matters while simultaneously telling them they have no say in how their work gets used. He wants the public to see a principled tech leader while the Pentagon sees a reliable vendor.
You can’t be both. And the moment someone tries, the community notes start writing themselves.
The Question He Won’t Answer
Altman tipped his hand with one revealing thought experiment from Monday evening. He asked himself how he’d feel “the day after an attack on the US or a new bioweapon we could have helped prevent.”
That’s not an ethics question. That’s a sales pitch. It’s the same argument every defense contractor has made since the Manhattan Project: the bad guys are coming, and you’ll wish you’d let us help.
It’s probably true. AI will play a critical role in national defense, and OpenAI’s technology is among the most advanced in the world. There are genuine arguments for government access to these tools.
But those arguments require honesty. You can’t sell the Pentagon access to your AI on Tuesday and tell the public you’re protecting their civil liberties on Monday. You can’t tell employees they have no vote and then post about “democratization” on social media.
Pick a lane, Sam. The Fourth Amendment defender or the defense contractor. The civil libertarian or the Pentagon’s AI guy. The principled CEO or the pragmatic businessman.
You cannot be all of them at once. And the internet now has the receipts.