Trump just handed the rest of the world the best argument for building their own AI

Over the weekend, Anthropic did something that would have been unthinkable a year ago. At the Trump administration’s request, the company pulled its newest and most powerful AI models—Fable 5 and Mythos 5—offline for all foreign nationals. That includes its own employees who aren’t US citizens.

Anthropic said it had little choice. The White House demanded a complete block on access, and when the government tells you to shut down, you shut down. No warning. No explanation beyond vague national security concerns. Just a sudden, sweeping order that effectively told the rest of the world: this technology is American, and you can’t have it.

Let me be clear about what this means in practice. These weren’t some experimental models nobody was using. Fable 5 and Mythos 5 were already deployed with safeguards limiting certain high-risk applications. They were state-of-the-art. Companies, researchers, and developers outside the US had built workflows around them. And now, poof—gone.

The irony here is thick enough to cut. For years, American AI companies have been selling the idea that frontier AI is a global public good. OpenAI talks about benefiting all of humanity. Anthropic’s whole brand is built on responsible, safe AI for everyone. But when push comes to shove, the US government can unilaterally decide that “everyone” actually means “Americans and whoever we feel like letting in today.”

This isn’t just an inconvenience. It’s a wake-up call for every country that’s been coasting on American AI infrastructure. If you’re building your startup, your research lab, or your national AI strategy on top of models that a foreign government can switch off overnight, you don’t own your own destiny. You’re renting it.

I’ve been saying this for a while, but incidents like this make the argument almost impossible to ignore. The EU, China, India, the UK—they’ve all been dabbling in domestic AI development, but the pace has been slow because American models are just better. Well, guess what? Better doesn’t mean much if you can’t access it when your geopolitical situation shifts.

The Trump administration’s action was characteristically blunt. No consultation with allies. No transition period. Just a demand and a deadline. Anthropic complied because defying the White House on a national security directive would have been corporate suicide. But the damage to trust is done.

I’ve talked to developers in Europe who are genuinely spooked. One told me they’re now prioritizing local models even if they’re less capable, because reliability matters more than peak performance when your business depends on it. That’s a completely rational response, and it’s going to accelerate the fragmentation of the AI ecosystem.

China, of course, is watching this closely. Beijing has been pushing hard on domestic AI, and every move like this from Washington makes their argument stronger: you can’t trust American tech. Build your own. The fact that China’s models come with their own set of control issues is a separate conversation, but the point stands.

The real tragedy here is that the safeguards Anthropic had already put in place were reasonable. The models were restricted from use in weapons development, surveillance, and other high-risk areas. That’s exactly the kind of responsible deployment the industry has been pushing for. But it wasn’t enough. The administration wanted total control, not responsible use.

What happens next? Other American AI companies are watching this precedent nervously. If Anthropic can be forced to shut down access overnight, so can OpenAI, Google, or Meta. The US government now has a playbook for weaponizing AI access in ways that go far beyond export controls on chips.

For the rest of the world, the message is unmistakable: American AI is a privilege, not a right. And privileges can be revoked. The countries that take this seriously will accelerate their own AI development, not because they want to compete with the US, but because they can’t afford to be dependent on a system that can be turned off at a politician’s whim.

I don’t know if the administration thought through the long-term consequences of this move. Probably not—it’s not exactly known for strategic thinking. But the effect is clear: they just made the most powerful case possible for non-American AI. And that case is going to resonate far beyond the tech press.