Anthropic has spent most of this week in a frantic game of digital whack-a-mole after the Trump administration abruptly ordered the company to cut off access to its newest AI models for all foreign nationals - including users physically inside the US and, awkwardly, the company's own employees. The result: Fable 5 and Mythos 5 were blocked for everyone, which is a bit like banning the color blue because some people might misuse it.
“To my knowledge, this is the first time US export controls have been used to control access to an AI model in this way,” said Hanna Dohmen, a senior research analyst at Georgetown University’s Center for Security and Emerging Technology. That's probably because export controls were originally designed for things you can drop on your foot, like weapons and hardware, not for nebulous digital services that live in the cloud and answer questions about recipes.
The Trump administration has not publicly explained the legal basis for the order, but Anthropic's website says the government cited “national security authorities” and concerns about a “jailbreak” potentially used by groups linked to China. (Anthropic also noted, with what one imagines was a weary sigh, that the jailbreak in question didn't actually let users bypass all safeguards.) Experts say the episode is unprecedented and exposes an uncertain stage in AI governance - or, as Andrew Reddie, a professor at UC Berkeley’s Goldman School of Public Policy, put it: “To say that this is an unsettled area of export control rule-making would be an understatement.”
The problem is that traditional export controls cover discrete things like software, source code, or 3D-printed gun files - items you can copy, download, or hand over. AI model weights (the core data that makes a model tick) were briefly controlled under Joe Biden, but the Trump administration abandoned that idea in its second term. The Anthropic order, however, targets something fuzzier: users don't receive model weights or source code; they just get chatbot answers. The “export” could be specific information produced by the models, or access itself - but remote access to cloud services is a known gap in current rules, one Congress is trying to close through legislation now moving through the Senate.
If Anthropic was targeted because Mythos and Fable are uniquely capable, the order raises obvious questions for every other frontier lab - OpenAI, Google, Meta, xAI, and the rest. If it was about specific safeguard issues, the government needs to outline what protection it considers sufficient. And if Anthropic was singled out because of its testy relationship with the Trump administration, the order becomes even harder to make sense of. Either way, experts say this is not a sustainable way to manage frontier AI, especially if the US wants to maintain its lead globally. Reddie summed it up: “In some ways, I think this episode makes clear the unsustainability of the existing governance regime.”
All of this points to the same problem: The Trump administration wants it both ways on AI - hands-off championing of American tech, but also ad hoc, unexplained orders that force a domestic champion to yank its frontier models. If Washington wants to control who can access powerful AI systems, it needs to say how, and give companies an actual chance to comply before launch. Otherwise, the US risks falling behind in the AI race, which would be a shame for a country that prides itself on being first at, well, everything.