Intelligence & Surveillance
The Pentagon Fired the AI Company That Wouldn't Spy on You. Then They Called OpenAI.
Last week the U.S. government cancelled a $200 million contract with an AI company because that company refused to help them spy on American citizens. Then they turned around and gave the contract to a company that said yes. Here's what that means for you.
How It Started
Back in 2025, the Department of Defense signed a deal with Anthropic, the company behind an AI assistant called Claude. At the time, Anthropic was clear about one thing: their technology could not be used for mass surveillance of people inside the United States, and it could not be used to power fully autonomous weapons systems. The Pentagon agreed. The contract was signed.
Then January came around.
Without much public fanfare, the Department of Defense came back and told Anthropic they wanted the restrictions removed. They wanted unrestricted access. No limits. No carve-outs. Full use of the technology for whatever the military decided it needed — including, specifically, domestic surveillance of American citizens.
Anthropic said no.
So the Pentagon fired them. They cancelled the contract and ordered every military contractor using Anthropic's products to stop immediately.
Within days, they had a new partner. OpenAI. The company that makes ChatGPT. The AI that's sitting on hundreds of millions of phones right now, including quite possibly yours.
When people found out, ChatGPT uninstalls jumped nearly 300%.
Good. But here's what most people aren't asking: why did this happen so fast? And what does it tell us about how your privacy actually works in this country?
Your Privacy Depends on a CEO's Mood
I want you to think carefully about what just happened here.
An AI company did the right thing. Their CEO drew a line and held it. And because of that, for a brief moment, a major government surveillance program hit a wall.
That's worth acknowledging. It takes real spine to say no to the Pentagon when $200 million is on the table.
But here's the problem. Your rights shouldn't depend on a CEO's personal ethics. They shouldn't hinge on whoever happens to be running a tech company this quarter, or what kind of mood they're in when the government comes calling, or whether their board decides the contract money is worth the public backlash.
"Your privacy is being decided in contract negotiations between giant tech companies and the U.S. government. You're not in the room. Nobody is asking you."
Even Anthropic's own CEO admitted this openly. He said in a recent interview that it's "Congress's job" to draw the legal lines around domestic AI surveillance. He described it plainly: the government buying bulk data on Americans, including their locations, personal information, and political affiliations, to build detailed profiles on ordinary citizens. He said the fact that this practice is currently legal means our laws haven't caught up with the technology.
He's right. But Congress isn't acting. A bill that would close the loophole allowing the government to buy your personal data without a warrant passed the House of Representatives in 2024. The Senate quietly buried it. Right now, as you read this, Congress is debating whether to reauthorize warrantless access to Americans' phone calls and texts under a law called Section 702. Don't hold your breath waiting for Washington to save you.
This Was Already Happening Before the OpenAI Deal
The OpenAI contract is not the beginning of this story. It's just the part that made the news.
The government has been building a surveillance infrastructure around your phone for years. And they've been doing it through purchasing rather than hacking, because buying data is perfectly legal and hacking requires a warrant. They figured out a long time ago that they didn't need to break into your phone. They could just buy everything it was already quietly producing.
-
Customs and Border Protection has been purchasing data directly from the online advertising industry to track the movements of people inside the United States — no warrant, no court order. -
Immigration and Customs Enforcement uses a tool that maps the precise locations of millions of devices using commercially purchased cell phone data. -
The Office of the Director of National Intelligence has been pushing to build a centralized marketplace where intelligence agencies can purchase commercially available personal data without going through the warrant process at all. -
The government holds massive contracts with data analytics companies that specialize in AI-powered analysis of enormous data sets, including detailed personal information on ordinary civilians.
None of this required a hack. None of it required a judge's signature. They just bought the data your phone was already generating and selling to advertisers every single day.
Now add AI to that picture. Add the ability to process billions of data points, cross-reference locations with names and faces and political affiliations, and build behavioral profiles at a scale no human analyst could ever manage manually. That is what the Pentagon wanted Anthropic's technology for. That is what OpenAI just agreed to help them do.
You Don't Need to Download Anything for This to Affect You
When the news broke, a lot of people had the same reaction: I'll just delete ChatGPT.
Do it. But understand that it doesn't fix the underlying problem.
Because AI is already running on your phone right now, and it has been for years. You didn't install it. It came with the device. It's baked into the operating system itself.
Siri has been listening on iPhones since 2011, always on, always waiting, processing your voice through Apple's servers. Hey Google runs continuously in the background on Android phones. Your keyboard uses on-device and cloud-based AI to predict what you're going to type next, and in most cases that prediction data flows back to a server somewhere. Your camera uses AI to recognize faces before you've even opened an app.
Deleting ChatGPT removes one layer. The operating system underneath it is still doing what it was built to do.
"They don't need to hack you. Your phone is already doing the work for them, every minute of every day."
The Only Real Answer Is What's Running Underneath
I've spent a long time studying how surveillance systems work. How data moves from a device to a server to an analyst to a file with your name on it. And the conclusion I keep arriving at is always the same: you cannot solve an operating system problem by managing your app list.
The fix has to go all the way down.
That's why I use a Ghost Phone.
A Ghost Phone is a Google Pixel device running GrapheneOS instead of standard Android. GrapheneOS is an open-source, privacy-hardened operating system that was built from scratch with one goal: remove the surveillance layer that ships inside every normal phone. There's no Google account required to set it up or use it. There are no background processes silently checking in with data brokers. The location harvesting that standard Android does as a default behavior is stripped out entirely.
You can still use it as a completely normal phone. Calls, texts, browser, apps — it all works. It just does all of it without quietly reporting your behavior to a network of companies and the agencies that buy from them.
The thing that struck me most about the Anthropic story isn't that the government asked. I've always known they'd ask. What struck me was the speed. One door closed on a Monday. By the end of the week they had a new partner and a new contract. No debate. No hearings. No public conversation about whether this is the kind of country we want to live in. Just a signature and a wire transfer.
Take Back Control of Your Phone
The Ghost Phone runs GrapheneOS. No tracking. No background data. No quiet deals with anyone.
Stay sharp.
Ed Warren — ZeroGhost™
This article draws on reporting and analysis by Matthew Guariglia at the Electronic Frontier Foundation. Original article: "The Anthropic-DOD Conflict: Privacy Protections Shouldn't Depend On the Decisions of a Few Powerful People" — EFF, March 3, 2026.
