Anthropic shut down its two most capable models three days after launching them, and it did not choose to. On June 12, 2026 the US government handed the company an export-control directive that forced it to disable Claude Fable 5 and Claude Mythos 5 for every customer worldwide. It is the first time a government has ordered a publicly deployed frontier AI model pulled from service, and it landed on the model creators had spent the week wiring into their pipelines.

What Happened

The directive reached Anthropic at 5:21pm ET on June 12 and took effect immediately. The order required Anthropic to suspend all access to Fable 5 and Mythos 5 "by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees." Because the company cannot reliably separate foreign nationals from US users in real time, the practical result was blunt. As Anthropic put it, "we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance."

Every other model stayed online. "Access to all other Anthropic models will not be affected," the company said, so Claude Opus 4.8 and the rest of the lineup keep running. CNBC and Fortune traced the order to the Commerce Department, which cited national-security concerns. Anthropic's own framing was narrower: "Our understanding is that the government believes it has become aware of a method of bypassing, or 'jailbreaking' Fable 5." Al Jazeera reported the order centered on barring access for all foreign nationals, the lever that forced the worldwide shutoff.

How Fable 5 Got Here: A Two-Week Timeline

The shutdown caps an unusually turbulent fortnight for what was meant to be Anthropic's showcase release.

Three days from a flagship launch to a global recall is a timeline the industry has not seen before, and it turned an ordinary upgrade cycle into a compliance emergency.

3D render of milestone markers on a track showing the Fable 5 launch-to-shutdown timeline
From flagship launch to global recall in three days.

Why a Jailbreak Triggered an Export Order

Export controls on AI have so far targeted chips and model weights crossing borders, not a switch that disables a hosted model for paying customers. The novelty here is the trigger. The government acted on a jailbreak finding: Anthropic says the directive followed a belief that someone had found a way to bypass Fable 5, and an administration official told Axios the Commerce Department moved after another company claimed it had jailbroken Mythos. (Anthropic's statement names Fable 5; the official cited Mythos, so the exact model behind the alarm is not fully settled in public reporting.)

The lever the government pulled was foreign-national access, which falls under export-control authority. Restricting a hosted model to US-only users in real time is not something Anthropic can do reliably on short notice, so the only way to comply at the speed the order demanded was to switch both models off entirely. That is how a "narrow potential jailbreak," in Anthropic's words, became a worldwide outage.

Anthropic disagrees with the call. The company said it does not believe "the finding of a narrow potential jailbreak should be cause for recalling a commercial model" used by hundreds of millions of people, and it warned the standard could stall new model launches across the industry. It says it considers the situation a misunderstanding and is working to restore access as soon as the order permits. NBC News reported the suspension holds until the government completes its security review, which officials said could take a few weeks, and Anthropic has not committed to a firm restore date.

3D render of a padlock clamped onto a model block, representing the forced shutdown
An export rule, not a safety recall, took the models offline.

Why It Matters: The First Government Takedown of a Frontier Model

Strip away the specifics and the precedent is the story. A jailbreak claim, surfaced by a competitor, was enough to pull a commercial frontier model from every customer overnight. No frontier lab has faced that before, and every one of them now has to plan launches against the possibility that a single security finding can trigger an export-control shutoff with no warning and no appeal window.

For the market, it reframes what a frontier model is. Treating Fable 5 and Mythos 5 as export-controlled technology, the way advanced chips are controlled, moves hosted AI closer to regulated dual-use goods than to ordinary software. That has knock-on effects for how labs document capabilities, how they gate foreign-national access from day one, and how much launch risk they are willing to carry. Anthropic keeping Opus 4.8 and everything else online also signals the controls are model-specific, not company-wide, which is small comfort to anyone who built this week on the two systems that are now dark.

3D render of an official stamp pressing a model cube, representing the first government-forced model takedown
The first time a government has pulled a live frontier model.

What Creators Should Do Now

If your workflow depended on Fable 5 or Mythos 5, the practical move is to repoint active jobs to Claude Opus 4.8, which remains available and handles most creative and coding work well. Keep your prompts and project files intact so you can resume on Fable 5 the moment access returns. For anything with a deadline, line up a second provider as a fallback rather than waiting on a restore date Anthropic has not committed to, and check Anthropic's status page before scheduling large batch runs this week. Apps and agents that hard-coded Fable 5 or Mythos 5 as their model will keep erroring until they are repointed, so audit any automation that names those models directly.

Frequently Asked Questions

Are Claude Fable 5 and Mythos 5 gone for good?

No firm answer yet. Anthropic calls the order a misunderstanding and says it is working to restore access as soon as possible, but it has not given a restore date, and the lockdown holds while the government's review runs.

Which Anthropic models still work?

Everything except Fable 5 and Mythos 5. Anthropic states that "access to all other Anthropic models will not be affected," so Claude Opus 4.8 and the rest of the lineup remain online.

Why did disabling foreign-national access shut the models off for everyone?

The directive bars access by any foreign national, inside or outside the US. Anthropic cannot reliably filter foreign nationals from US users on a hosted model in real time, so a full shutoff was the only way to comply quickly.

Who found the jailbreak?

Public reporting is not fully settled. Anthropic's statement says the government believes there is a way to jailbreak Fable 5; an administration official told Axios the Commerce Department acted after another company claimed it had bypassed Mythos.

Is this the same as the earlier Fable 5 guardrail apology?

No. The June 11 apology was about a hidden safeguard Anthropic built into Fable 5 that degraded some development work. The June 12 shutdown is a separate, government-ordered export-control action.

Does my data or output disappear?

The models returning errors does not delete your saved prompts or project files. Keep them so you can resume if and when Fable 5 and Mythos 5 come back.

For background on what these models offered and the events leading up to the shutdown, see our coverage of what Fable 5 gives creators versus Opus 4.8, the invisible-guardrail apology, and the report on Fable 5 access through Amazon Bedrock.