US Order Forces Anthropic to Disable Fable 5 and Mythos 5 Models

For a few hours on Friday evening, the most capable commercial models Anthropic has ever shipped were live. Then a letter arrived from the US government, and by night the company had switched them off, for everyone.

Anthropic said in a statement that federal agencies had ordered it to block all foreign nationals from using Fable 5 and Mythos 5, its newest frontier systems, on national security grounds. Unable to fence off only non-citizens at short notice, the company disabled the two models for its entire customer base instead.

The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance, the company said. The order landed at 5:21pm Friday, it added, and the letter did not spell out the specific concern.

Mythos 5 is the part that worries Washington. The model is unusually good at finding software vulnerabilities, including flaws that have sat undiscovered in code for decades, and US agencies and a handful of corporate partners have been using it to patch their own systems before anyone else finds the holes. The same skill, pointed the other way, is a cyberweapon. Fable 5, released only this week and built on the same underlying technology, ships with those cybersecurity and biotechnology capabilities walled off; Mythos 5, the full version, was never public.

Anthropic says it is working from partial information. After reviewing what it believes triggered the directive, the company concluded the government had found a narrow way to make the model review specific program code and fix errors, a capability that rivals such as OpenAI's GPT-5.5 also have, it noted. The reported issue, in its telling, is a limited jailbreak, not a new weapon.

That is the crux of the objection. Anthropic argued that pulling a product used by hundreds of millions of people over a narrow, already-tested flaw sets a standard that, applied across the industry, would freeze new releases everywhere. It is an awkward position for a firm that, only days earlier, had urged the world's leading labs to consider pausing the development of ever-faster systems for fear humans could lose control.

It also fits a wider pattern. Washington has spent the past year tightening who can touch its most advanced computing, from chip sales to China to the models themselves, even as consumer adoption of AI keeps breaking records. Friday's order pulls the model layer into the same export-control logic that has long governed semiconductors.

The reaction in Europe was less about Anthropic and more about the plug. Politicians across the continent read the episode as proof of how exposed they are. Europe cannot settle for being an open market dependent on technologies designed, funded, and controlled elsewhere, wrote Benjamin Haddad, France's minister delegate for Europe, who called the move an accelerator of the geopolitical battle over AI.

Geert Wilders, the Dutch politician, put it more plainly on X: I want my #Anthropic Claude Fable 5 back! He added that AI is, more and more, a matter of national sovereignty. The takeaway echoing through European capitals, per Euronews: a country that rents its most powerful technology from a foreign government can have it switched off overnight.

For now the models stay dark while Anthropic and the government work through what, exactly, the concern is. The company has the bigger argument it would rather be having — that frontier AI is getting powerful enough to need brakes — and an immediate one it did not choose, about who holds the off switch. On Friday, the answer was not in doubt.