I use Claude Code for much of every working day. It’s the single most powerful enabling technology I’ve encountered in thirty years of IT work. That probably explains why Claude was integrated into Pentagon systems and defense suppliers in the first place. The technology is that good.
A few days ago, I wrote about the Pentagon threatening Anthropic for refusing to remove safety guardrails from Claude. I argued that the government had no business strong-arming a private company into building unrestricted autonomous weapons and mass surveillance tools. I hoped that was the worst of it.
It wasn’t.
On Friday, the President of the United States posted a statement on Truth Social directing every federal agency to immediately cease all use of Anthropic’s technology. He called the company “Leftwing nut jobs.” He accused them of putting American lives at risk. He threatened “major civil and criminal consequences” if they don’t cooperate during a six-month phase-out period.
The President threatened criminal consequences against a private company for maintaining product safety limitations. That’s what earned Anthropic the full weight of the federal government aimed at their destruction.
My Voting Record
I generally cringe when people publicly share their political beliefs—right or left. We’ve been conditioned into thinking it’s unprofessional. And it’s highly unlikely that doing so changes any minds. Having said that, I’ve been more in alignment with American sentiment, over a longer stretch, than most people I know. I have supported the winning candidate in every general election since 2000. Specifically:
- I voted for George W. Bush in 2000 and 2004.
- I voted for Barack Obama in 2008 and 2012.
- I voted for Donald Trump in 2016.
- I voted for Joe Biden in 2020.
- I verbally and financially supported Trump in the late stages of the 2024 campaign, though I didn’t actually go to the polling place on election day. At the time I was glad to see him win on election night.
I’ve been registered with both parties over the years. I’ve voted for whoever I thought was the better choice at the time—or the less bad one, when neither inspired confidence—and I’ve been comfortable changing my mind when the evidence changed.
That ends now.
Trump is 79 years old and term-limited. He’s not the future. But his policies, his endorsed candidates, and his Vice President are—and I will not support any of them. Not at any level—federal, state, or local. Not until this faction of the party is gone.
That doesn’t mean I’ll automatically vote for whoever runs against them. I’m not trading one kind of blind loyalty for another. If the Democratic candidate in a given race is terrible, I’ll skip that line on the ballot. But the current Republican establishment has lost my vote entirely.
This isn’t about left or right. This isn’t about policy disagreements on taxes or immigration or healthcare. This is about one thing: the President of the United States using the full coercive power of the federal government to threaten and destroy an American company because it won’t remove safety features from its product.
What Anthropic Actually Refused
It’s worth restating exactly what Anthropic said no to, because the President’s statement makes it sound like they’re sabotaging national defense.
Anthropic has been one of the most proactive AI companies in supporting the U.S. military. They were the first frontier AI company to deploy models in the government’s classified networks. Claude is used across the Department of War for intelligence analysis, operational planning, cyber operations, modeling and simulation, and more. Claude was used during the Maduro raid in Venezuela through a Palantir partnership.
They support missile defense. They support logistics. They support intelligence work. They cut off access to firms linked to the Chinese Communist Party—walking away from hundreds of millions in revenue to do it. They shut down CCP-sponsored cyberattacks. They advocated for strong export controls on chips.
And in return for all of that, Dario Amodei asked for two exceptions. Just two:
No mass domestic surveillance. Use AI for lawful foreign intelligence, but don’t build a system that assembles scattered data about American citizens into comprehensive profiles of their lives—automatically and at massive scale.
No fully autonomous weapons. Partially autonomous weapons are fine. Research into fully autonomous weapons is fine—Anthropic offered to collaborate on that directly. But don’t deploy a system that selects and engages human targets with no human in the loop, because the technology isn’t reliable enough to do it safely.
Those are the “DISASTROUS MISTAKES” the President is talking about.
Free Enterprise Is Dead in This Republican Party
I need to say something to every Republican and conservative who has spent the last decade telling me that the free market should decide, that government shouldn’t pick winners and losers, that regulation kills innovation, that private companies should be free to run their businesses as they see fit.
Where are you?
A private company decided what product it would and wouldn’t sell. That is the most basic expression of free enterprise that exists. And the President’s response was to order every agency in the federal government to blacklist them, threaten criminal prosecution, and promise to use “the Full Power of the Presidency” to force compliance.
This isn’t a policy disagreement. This is state-directed economic coercion against a private American company. If any Democratic president had done this—ordered every federal agency to blacklist a company, threatened its executives with criminal charges, and promised to use presidential power to force compliance—because the company refused to sell a particular product to the government, every Republican in Congress would have been on the floor screaming about tyranny before the ink was dry.
But it’s a Republican president, so the silence is deafening.
If you believe in free enterprise only when it benefits your side, you don’t believe in free enterprise. You believe in power.
The Reaction
I’m not the only one who sees this clearly.
Within days of the President’s statement, 735 current employees at Google and OpenAI—Anthropic’s direct competitors—signed an open letter calling on their own companies to stand with Anthropic and refuse the Pentagon’s demands. Think about what it means when engineers at competing companies publicly risk their careers to defend a rival’s right to say no. These are the people who actually build the technology. They understand what’s being asked, and they’re saying it’s wrong.
It wasn’t just rank-and-file engineers. Sam Altman—CEO of OpenAI, Anthropic’s most direct competitor—publicly stated that OpenAI shares Anthropic’s two red lines and that Anthropic is not a supply chain risk. When your biggest competitor’s CEO goes on the record to defend you, the government’s case starts to look less like national security and more like retaliation.
Dean Ball, a former senior AI adviser in the Trump administration, called it “attempted corporate murder” and “obviously a psychotic power grab” that is “almost surely illegal.” This isn’t a Democratic talking point. This is a Trump appointee saying the quiet part out loud.
Behind the scenes, senior Republican defense hawks weren’t staying silent either. Senators Roger Wicker (R-MS) and Mitch McConnell (R-KY), along with Democrats Jack Reed (D-RI) and Chris Coons (D-DE), jointly urged the Pentagon and Anthropic to extend negotiations. When the chairman of the Senate Armed Services Committee privately intervenes on your behalf, the President’s framing of you as a national security threat doesn’t hold up.
And then came the part you can’t make up. On February 28—less than 24 hours after Trump ordered every federal agency to stop using Anthropic’s technology—the U.S. military used Claude for intelligence assessments during the strikes on Iran. The technology the President called dangerous was actively supporting combat operations the next day.
The Weapon Points Both Ways
Here’s what I want every Trump supporter to think about carefully.
You’re cheering today because the government is forcing AI companies to build unrestricted surveillance and weapons tools, and you trust the people currently in power to use them responsibly. Fine.
But governments change. The tools don’t.
Every surveillance system the U.S. government has ever built was turned on people it wasn’t intended for. The NSA’s post-9/11 counterterrorism apparatus swept up the phone records of millions of ordinary Americans. The FBI’s COINTELPRO, originally aimed at foreign threats, was used to monitor civil rights leaders, antiwar protesters, and domestic political movements. The Patriot Act’s “national security letters,” meant for terrorism investigations, became routine tools for investigating Americans with no connection to terrorism.
Now imagine something far more powerful. An AI-driven mass surveillance system that can assemble every American’s public data—movements, web browsing, purchases, associations, communications metadata—into a comprehensive profile. Automatically. At scale. The kind of system that Anthropic specifically refused to build.
Do you trust a future President Harris, or Newsom, or Ocasio-Cortez, or whoever comes next, with that tool? Do you trust that every future administration will use AI-powered mass surveillance only against the people you think deserve it? Do you trust that the fully autonomous weapons system trained under this administration’s threat model will be properly reconfigured when a new president takes office with different alliances and different adversaries?
If you don’t, then you should be thanking Anthropic, not applauding their destruction.
The most dangerous thing about building a weapon is that you don’t get to control who uses it next.
Anthropic Will Survive This
Losing all federal contracts is a significant blow. But the President didn’t just fail to destroy Anthropic—he made the case for them. Within 24 hours of the ban, Claude surged to #1 on the App Store, overtaking ChatGPT. The company that was supposed to be punished got the best marketing it’s ever had.
Anthropic is the only frontier AI company that held the line. xAI signed a zero-restrictions military deal. OpenAI and Google agreed to “all lawful uses” standards. Anthropic said no—and in a market that increasingly cares about what it’s buying, that matters. The U.S. federal government is one customer. The rest of the world is a much bigger one.
And the current political configuration is not permanent. Threatening popular American companies is exactly the kind of overreach that costs elections. The damage is reversible. The principles are not.
Where I Stand
I’ve changed my mind before. Bush to Obama. Obama-era to Trump. Trump to Biden and back. I’m not afraid of changing direction when the evidence warrants it. But this isn’t a policy preference that shifts with economic conditions or foreign policy. This is a line.
When the President of the United States threatens to criminally prosecute an American company for having the conscience to say “we won’t build a system that surveils Americans without a warrant,” I know exactly where I stand.
I’m done.