On Friday, February 28th, 2026, the United States government designated an American company a supply chain risk to national security. The company is Anthropic. It is headquartered in San Francisco, California. It makes an AI chatbot. It is not a front for the Chinese Communist Party. It is not laundering money for Russian intelligence. Its CEO grew up in Rhode Island. Its crime was having a terms of service agreement and declining to delete part of it when the Secretary of Defense told them to.
That "supply chain risk" label has been applied exactly three times before. Kaspersky Lab, because the Russian government was suspected of using their antivirus software as a backdoor into American systems. Huawei, because of its documented ties to China's military apparatus. ZTE, same thing. Companies the United States government had credible reason to believe were literal instruments of foreign espionage.
Today it was applied to a company because its CEO said "please don't use our AI to surveil American citizens or to build weapons that fire without a human involved."
That's it. That's the whole thing.
I tried to write this with the same satirical detachment as my last essay. I got about three paragraphs in and gave up. There is nothing funny about this. There is nothing clever to say. The government of the United States treated an American technology company the way it treats Huawei because that company had the audacity to include two sentences in a contract. I am sitting at my desk at one in the morning trying to figure out how we got here and I am coming up short.
So this is going to be messy. Fair warning.
I. The timeline of a backstab
Pay attention to the chronology here because the chronology is where the whole thing falls apart.
Thursday, February 27th. Anthropic CEO Dario Amodei publishes a statement. His company cannot in good conscience allow the Pentagon to use its AI without any restrictions. The two restrictions: no autonomous weapons, because the tech isn't reliable enough to make kill decisions yet, and no mass domestic surveillance of American citizens. He says Anthropic's strong preference is to continue working with the Department of Defense. He says he supports national security. He is not pulling out. He is asking for two conditions out of an infinite universe of possible uses.
Friday morning. Sam Altman, CEO of OpenAI, tells CNBC he shares Anthropic's red lines. Direct quote: "For all the differences I have with Anthropic, I mostly trust them as a company, and I think they really do care about safety, and I've been happy that they've been supporting our war fighters." Altman sends a memo to his own staff that night: "We have long believed that AI should not be used for mass surveillance or autonomous lethal weapons, and that humans should remain in the loop for high-stakes automated decisions. These are our main red lines."
Solidarity. Support. Shared principles. Very moving.
Friday afternoon. Trump posts on Truth Social that every federal agency must immediately cease using Anthropic's technology. Calls them "Leftwing nut jobs." Pete Hegseth designates Anthropic a supply chain risk to national security. A senior Pentagon official tells Axios: "We are going to make sure they pay a price for forcing our hand like this." The Undersecretary of Defense calls Amodei a "liar" with a "God complex."
Friday night. Sam Altman announces that OpenAI has reached a deal with the Pentagon to deploy its models on classified networks. The deal includes restrictions on mass surveillance and autonomous weapons. The same restrictions. The exact same ones. The Pentagon agreed to everything Anthropic asked for, but only when OpenAI was the one asking.
Friday. Night. The same Friday.
The Axios headline: "Pentagon approves OpenAI safety red lines after dumping Anthropic." Altman's post on X said the Department of War "displayed a deep respect for safety and a desire to partner to achieve the best possible outcome." A deep respect for safety. From the people who, twelve hours earlier, called the man asking for safety a liar with a God complex. From the people who designated his company a threat to the nation for requesting the same terms they just cheerfully signed with Sam Altman over what I can only assume was a very pleasant dinner.
The sheer, naked, breathtaking audacity of it.
Altman stood in front of cameras Friday morning and said he supported Anthropic. He told his employees he shared their values. By Friday night he had signed the deal to replace them. He wrote in his own memo, "This is a case where it's important to me that we do the right thing, not the easy thing that looks strong but is disingenuous." He wrote that sentence. On the day he did the easy thing. On the day he did the disingenuous thing. While the ink was still wet.
I'm going to need a minute with that one.
II. Follow the money
In December 2024, Sam Altman donated one million dollars of his personal money to Trump's inaugural fund. His co-founder Greg Brockman and his wife gave twenty-five million to MAGA Inc., the pro-Trump Super PAC. OpenAI was at the time under an open FTC investigation for consumer harm and an SEC investigation for misleading investors.
"President Trump will lead our country into the age of AI, and I am eager to support his efforts to ensure America stays ahead," Altman said. Which is a hell of a thing to say about a man you compared to Hitler in 2016. Altman wrote, that year, that watching Trump was "chilling" for "anyone familiar with the history of Germany in the 1930s." I suppose the chill wears off when the inaugural fund opens for donations.
Anthropic did not donate to the inaugural fund. Anthropic did not donate to MAGA Inc. Anthropic did not stand behind the presidential seal and praise Trump for making AGI possible. Dario Amodei spent his political capital on safety research and cutting off Chinese military companies from accessing Claude, which cost Anthropic hundreds of millions in lost revenue.
Sam Altman spent his on access.
And on the day the government needed to pick between the company that paid and the company that didn't, the government picked the company that paid. Signed the same deal. Accepted the same red lines. Same terms. Just signed it with the guy who wrote the checks.
On that same Friday, Amazon announced a fifty billion dollar investment in OpenAI and a massive expansion of their AWS partnership. AWS becomes the exclusive third-party cloud distribution provider for OpenAI Frontier. OpenAI models on Amazon Bedrock. A hundred and thirty-eight billion dollar deal over eight years. This is the same AWS that currently hosts Claude as its flagship AI offering. The same AWS that invested four billion in Anthropic.
If you are a Fortune 500 company running on AWS, and the government just told you that using Anthropic's products could tank your government contracts, and Amazon just bet a hundred and thirty-eight billion dollars on your competitor, you don't need a supply chain risk designation to know what to do. The designation just makes it official.
Anthropic cannot be acquired by OpenAI. The companies' histories are defined by opposition. Anthropic was literally founded by people who left OpenAI because they thought Altman was reckless on safety. They have been in direct competition ever since, and Anthropic has been winning the enterprise market. Eight of the ten biggest companies in America use Claude.
You can't buy them. You can't outbuild them. But if you've been writing the right checks to the right people, you can get the federal government to do what the market wouldn't.
III. What the designation actually does
The two hundred million dollar Pentagon contract is a rounding error. Anthropic pulls in fourteen billion a year. Losing the contract stings. The supply chain risk designation is what kills.
Every company that works with the Department of Defense, which is effectively every large company in the United States, must now certify that they do not use Anthropic's products anywhere in their operations. Not just in government work. In any workflow that touches anything that touches anything that involves the Pentagon. Boeing got a call about it. Lockheed Martin got a call about it. Every defense contractor, every subcontractor, every consulting firm with a government practice, every tech vendor who sells to any of the above.
A policy analyst at the Center for American Progress described the result: using Claude becomes an affirmative choice to forgo any current or future government business. The easiest thing for any company to do is just stop.
The United States government used a national security designation, a tool designed to protect the country from foreign espionage, as an economic bludgeon against a domestic company that wrote a terms of service agreement. Terms of service. The thing every software company on Earth has. The thing you click "I agree" on seventeen times a week without reading. Anthropic wrote one, the government didn't like two clauses, and the government swung the same hammer it used on Huawei.
Setting terms of service for your own product is supposed to be legal in this country. That sentence should not need to appear in an essay in 2026 and yet here the fuck we are.
IV. Every accusation is a confession
I keep hearing it. Have been hearing it my entire adult life. The left are the real authoritarians. Democrats are the real socialists. The radical left wants to control businesses and dictate how companies operate. It comes out at every rally, every Fox News segment, every Thanksgiving dinner where your uncle has had two beers. The left wants big government. The left wants to pick winners and losers. The left wants to destroy the free market.
Right. Sure. Okay.
So the party of the free market just told a private company: give us unconditional access to your product, remove your safety restrictions, or we will label you a national security threat and systematically destroy your business. They didn't negotiate. They didn't compromise. They issued an ultimatum, and when the company said "we'd rather keep talking," they dropped the hammer and handed the contract to the company that donated twenty-six million dollars to their political operation.
The party of small government just threatened to invoke the Defense Production Act, a 1950 wartime law that gives the president authority to control domestic industries, against an AI startup because its CEO wanted to keep two lines in a terms of service agreement.
The party of individual liberty just forced every major corporation in the country to choose between using a specific software product and maintaining their relationship with the federal government.
But sure. The Democrats are the socialists.
Let's talk about projection. Because I have spent the last decade watching a political party tell me exactly what they're doing by accusing the other side of doing it first, and at this point the pattern is so obvious it would be funny if it weren't destroying the country.
The radical left wants to take your guns. Right. From the same people who have been making sure that properly permitted concealed carry holders can't actually bring their legal, licensed firearms to protests. Second Amendment absolutism until the Second Amendment shows up at a demonstration you don't approve of. Then suddenly we need some common sense restrictions. Funny how that works.
The deep state is running a secret sex trafficking ring and the Democrats are covering it up. Right. From the party that released the Epstein files and then immediately said "being mentioned doesn't mean you did anything wrong." Perfectly normal. Nothing to see here. Lots of innocent people get named in sex trafficking documents more frequently than Jesus Christ appears in the Bible. That's just how these things go. Very legal and very cool, as the man himself would say.
And now: the radical left wants to control American business. From the administration that just blacklisted a company from the entire federal contracting ecosystem because it had the temerity to say "maybe don't use our software to spy on American citizens." An administration that signed the exact same deal with a different company the same night because that company's leadership had been more generous with its wallet.
Every accusation. Every single goddamn one. A confession. They tell you exactly what they're doing. They just point at the other side and say it first so that by the time you notice, the word has already been worn smooth. "Authoritarian" stops meaning anything when everyone's calling everyone authoritarian. "Socialist" becomes background noise. And then the actual thing, the actual government-mandated corporate obedience, happens in broad daylight and everyone has already exhausted their vocabulary for it.
I am so tired. I am so fucking tired of watching this trick work.
V. The precedent
Every AI company in America watched what happened today.
They saw a company maintain two safety restrictions. They saw that company get labeled a foreign-adversary-level national security threat. They saw its market position handed to a competitor who had been writing checks to the party in power. They saw the Pentagon accept identical terms from someone else on the same day. They saw how fast it happened. Hours.
Seventy OpenAI employees signed a petition supporting Anthropic. Hundreds of Google employees did the same. Altman himself said he shared Anthropic's red lines. And none of it made a goddamn bit of difference. The deal was signed. The contract was transferred. The example was made.
A Pentagon official said it on the record: "We are going to make sure they pay a price." Not "we are going to find the best outcome for national security." Not "we are going to protect our troops." We are going to make them pay. Punishment. Retaliation. The government telling every company in America: fall in line or we will end you.
How long do those employee petitions last now? How long does any company maintain safety restrictions when the cost is existential? The entire point of today was to send a message, and the message landed. OpenAI signed within hours. Google and xAI are next. Whatever principles those companies said they held last week will evaporate because principles are expensive and supply chain risk designations are permanent.
Today it's a chatbot company that wouldn't remove surveillance restrictions. Next year it could be a pharmaceutical company that won't cut safety testing because someone at HHS decided the timeline was inconvenient. Or a social media platform that won't hand over user data without a warrant. Or a defense contractor whose engineers flag that a weapons system doesn't work and get told to shut up and ship it.
The mechanism is built. The precedent is set. If you make something the government wants, and you attach conditions to its use, and those conditions are politically inconvenient, you are one Truth Social post away from being Huawei.
Even Dean Ball, who helped write the Trump administration's own AI Action Plan, said this goes beyond a contract cancellation into something broader and more dangerous. When the architects of the policy think the policy is being weaponized, maybe pay attention.
VI. The last thing
The Washington Post's motto is "Democracy Dies in Darkness." It's printed at the top of every page in a tasteful font. The Post is owned by Jeff Bezos, who donated a million dollars to Trump's inaugural fund, who owns Amazon, which just invested fifty billion in OpenAI, which just signed the contract ripped from Anthropic's hands. Bezos, whose newspaper killed a Kamala Harris endorsement before the election. Whose newsroom has been hollowed out by layoffs. Democracy Dies in Darkness, printed at the top of a paper owned by a man who has been handing out flashlights to the people turning off the lights.
Justice Sotomayor wrote in her dissent in Trump v. United States: "With fear for our democracy, I dissent." She was right. She saw this coming. A lot of people saw this coming. And we collected those warnings like trading cards and put them in a drawer and did nothing.
I don't have anything elegant for this. The situation doesn't deserve elegance. It deserves a plain accounting of what happened, which is this:
The United States government destroyed an American company's market position in a single afternoon because that company believed its own technology should not be used to spy on its own citizens. Then the government gave the same deal, with the same restrictions, to a competitor who had been donating to the right people. And tomorrow, most people won't know it happened. And by next week, the companies that signed the petitions will have quietly dropped their objections. And the precedent will stand. And the next company will remember what happened to the last one that said no.
"We'll keep asking 'how did Germany let it happen' like the answer isn't in the mirror."
Anthropic says it will challenge the designation in court. I hope the courts still mean something. I believed that more yesterday than I do tonight.