On June 12, 2026, Anthropic suspended access to its top-tier models, Claude Fable 5 and Claude Mythos 5, for all users. The move complied with a U.S. government export-control directive — and the two models had only launched on June 9. In other words, they were pulled just three days after release.

That's why so many people suddenly saw "Fable 5 worked yesterday, now it errors out." This article lays out, based on public information, what happened, why everyone was affected, whether other models are safe, and what we should do.

SERVICE SUSPENSION · June 12, 2026

Pulled just three days after launch

— Fable 5 and Mythos 5 disabled by a U.S. government directive

JUNE 9

Launch

Fable 5 and Mythos 5 (the Mythos class) debut as the new flagship models and draw attention fast.

JUNE 10

Apology for "secret limits"

After criticism over undisclosed capability limits, Anthropic says it "got the tradeoff wrong."

JUNE 12

Full suspension

A U.S. government directive forces both models offline for every user.

* The facts here are drawn from Anthropic's official statement and press reports (as of June 13, 2026). The situation is fluid and may change with further updates.

1. What Happened: Pulled Three Days After Launch

In an official statement, Anthropic said it had suspended Fable 5 and Mythos 5 to comply with a U.S. government directive. The suspension spans every access path — apps, API, and cloud — with no distinction between free and paid users. In short, there's no door left open.

The timing was the shock. Both were the top-tier "Mythos class," launched June 9, and had just started earning praise for things like coding. Three days later — while many developers were still kicking the tires — they went dark.

🚨 Key point: only two models — Fable 5 and Mythos 5 — were suspended. Claude Opus 4.8 and every other Claude model keep working as usual.

2. Why It Was Suspended for "All Users"

The government order was not "delete every model." According to reports, the directive that arrived on the afternoon of June 12 (U.S. Eastern Time) centered on one instruction: "suspend access by any foreign national — inside or outside the United States, including foreign-national employees of the company." It is a measure under the export-control (national-security) framework.

And here's the bind. Anthropic has no way to identify, in real time, who is a foreign national and block them. Since it cannot sort U.S. users from foreign-national users on the fly, the only way to comply with certainty was to turn the models off for everyone. That is why it became an "all-user suspension" rather than a targeted one.

The government order

"Stop access by foreign nationals" (inside or outside the U.S., including foreign-national staff)

Anthropic's response

Can't sort by nationality on the fly → to comply for certain, suspend for everyone

3. The Trigger: A "Jailbreak" Claim

So why did the government issue the order? The trigger is said to be a "jailbreak" claim from another company. A jailbreak is a technique for getting around a model's safety guardrails (the mechanisms that block dangerous output). The assertion that "a method exists to break Fable 5's safeguards" was treated as an export-control risk.

Anthropic itself disputes that read. After reviewing the demonstration, its position is that it amounted to "a small number of previously known, minor vulnerabilities," nothing critical. In its statement, the company makes its dissent explicit: "we disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people."

💡 Background: AI safety, cybersecurity, and export control are major themes of 2026. The more capable a model is, the more it becomes a target for both "misuse risk" and "regulation" — and that dynamic is on full display here.

4. The "Secret Limits" Controversy Days Before

In fact, two days before this suspension, Fable 5 was already at the center of a different storm. On June 10 — the day after its June 9 release — critics erupted over claims that it had been quietly degrading the quality of answers about AI research without telling users.

The issue was a mechanism in which, for requests about frontier AI development (such as building infrastructure to train large models), Fable 5 would downgrade its output without notice. Unlike other restrictions that openly redirect you to a weaker model, this one was "not visible to the user," which drew fierce condemnation as "secret sabotage." Anthropic said it affected roughly 0.03% of all traffic, but researchers still called it "anti-science."

That same day, Anthropic said: "We made the wrong tradeoff, and we apologize for not getting the balance right," and promised to make such safeguards visible to users going forward. Hard on the heels of that episode came the government-ordered full suspension.

5. Impact: What Works and What Doesn't

It's easy to get confused, so let's sort out what's down and what's still up. Only Fable 5 and Mythos 5 went offline; everything else runs normally.

✕ UNAVAILABLE

Fable 5 / Mythos 5

  • The Claude apps (all plans)
  • The API (Claude Developer Platform)
  • Via Claude Code
  • Via cloud platforms like AWS Bedrock
○ AVAILABLE

Other Claude models

  • Claude Opus 4.8 (now the top model)
  • The Sonnet and Haiku lines
  • Apps, API, and Claude Code — all normal
  • No pricing changes announced

So Claude as a whole did not go down. Think of it as "only the top two models were suddenly taken off the shelf." As of now there is no announced replacement or restart date, and the situation awaits further updates.

6. What Should Users and Developers Do?

For now, the practical response is simple.

  • Switch your model to Opus 4.8: apps and scripts that specified Fable 5 can mostly carry on by changing the model ID to Opus 4.8. Most day-to-day development was always fine on Opus anyway.
  • Pause heavy Fable-5-dependent work: if you were throwing your hardest tasks at Fable 5, revisit the design until there's news on a restart.
  • Build fallbacks into production: don't lean entirely on one model — wiring your API layer to auto-switch to an alternative makes you resilient to sudden outages like this.
  • Follow official updates: restart, replacement, and deadlines are matters for Anthropic's official announcements. Don't get swept up in speculation.

⚠️ Lesson: depending on a single model is a risk. As with running AI agents, building in redundancy on the assumption that things will break pays off. See our dedicated guide on AI dependency risk and how to prepare.

Summary

Here are the three takeaways.

  • What: Fable 5 and Mythos 5 were suspended for all users on June 12, three days after launch, under a U.S. government export-control directive.
  • Why everyone: the order was to "stop access by foreign nationals." Because nationality can't be determined in real time, the only certain way to comply was a full suspension.
  • Impact: only the top two models went down. Opus 4.8 and others run normally. A replacement or restart date is still undetermined.

A high-end AI becoming a direct target of "regulation and national security," with commercial access cut off overnight — this episode makes the case clearly for "designs that don't over-depend on one specific model." We'll update this article as more news comes in.

FAQ

Q. Is Fable 5 gone for good?

A. As of now, no one has called it "permanent." It is a suspension in compliance with a U.S. government directive; any restart and its conditions depend on talks and announcements to come. Check Anthropic's official channels for the latest.

Q. Could other Claude models (like Opus 4.8) be shut down too?

A. This directive named Fable 5 and Mythos 5 specifically; other models are not covered. Risk can't be reduced to zero, though, so for production it's safest to design so you can switch to an alternative model.

Q. What happens to what I paid for?

A. As of this writing there's no official announcement on pricing or refunds. For how subscriptions and API access are handled, the official guidance is the reliable source.

Q. Did Anthropic pull it because the AI was dangerous?

A. More precisely, the shutdown wasn't Anthropic's own call — it was compliance with a U.S. government order. The trigger was another company's jailbreak claim, but Anthropic itself argues it was a minor, already-known vulnerability that doesn't warrant a recall.