If you opened Claude Code today and hit a wall that reads “There’s an issue with the selected model (claude-fable-5[1m]). It may not exist or you may not have access to it. Run /model to pick a different model,” you are not alone, and nothing is broken on your end.
On June 12, 2026, the US government ordered Anthropic to suspend access to its two most powerful models, Claude Fable 5 and Claude Mythos 5, for every customer in the world. Anthropic complied the same day. The error you are seeing is the result.
This happened just three days after Fable 5 launched as the most powerful model Anthropic had ever released to the public. It went from headline launch to government takedown in 72 hours, and it is the first time a US AI lab has been forced to pull a generally available model off the market on national security grounds.
Here is the full picture: what the error actually means, what you should do right now, why the government stepped in, what Anthropic says about it, and when the model might come back.
What the Error Message Means
The error is not a bug, a billing problem, or anything tied to your install. Anthropic flipped the model off at the server level, so every client that tries to call it gets rejected.
You may see it in a few different forms depending on where you are working:
- In Claude Code: “There’s an issue with the selected model (claude-fable-5[1m]). It may not exist or you may not have access to it. Run /model to pick a different model.”
- A shorter “Fable is not available” or “Claude Fable 5 is not available. Please use Opus 4.8.”
- On the API: “Invalid or Inaccessible Model claude-fable-5”
All of these point to the same cause. The model identifiers that no longer resolve include claude-fable-5, the one million token variant claude-fable-5[1m], and the AWS Bedrock IDs anthropic.claude-fable-5 and global.anthropic.claude-fable-5. The [1m] tag is simply the long context version of the same model, the one tuned for the full one million token window.
GitHub filled with near identical reports within hours of the suspension, which is the clearest sign this is a platform wide change and not an account specific one.
What To Do Right Now
The fix takes a few seconds. You do not need to reinstall anything, log out, rotate API keys, change your network, or contact support. The only thing that changed is which model answers you.
- In Claude Code, run
/modeland pick another model, such as Claude Opus 4.8, Sonnet 4.6, or Haiku 4.5. - On the API, swap the model string in your request from
claude-fable-5toclaude-opus-4-8or another available ID. - In any app that lets you choose a model, select anything that is not Fable 5 or Mythos 5.
Every other Anthropic model is running normally. Opus 4.8 is the closest substitute, and for many tasks it is the model Fable 5 was already falling back to behind the scenes.
What Actually Happened
Anthropic received the directive on June 12, 2026 at 5:21pm ET and acted on it immediately, suspending both models the same evening. The company published a statement the following day explaining the move.
The order came through export control authority. According to reporting on the directive, it was issued by Commerce Secretary Howard Lutnick to Anthropic CEO Dario Amodei and drafted with the Commerce Department’s Bureau of Industry and Security, the same arm of government that polices exports of sensitive technology.
Two details make the scope unusually broad. First, it covers both the safeguarded public model, Fable 5, and the unrestricted partner model, Mythos 5. Second, it is not aimed at any single country. The directive reportedly applies to “any foreign national, whether inside or outside the United States,” which in practice means the models were switched off for everyone, US citizens included, because Anthropic cannot cleanly separate users that way at the model level.
So this is not a regional block or a sanctions list. It is a full global suspension of two specific models while everything else Anthropic offers stays online.
Why the Government Stepped In
The stated reason is national security. According to Anthropic’s own account, the government cited awareness of a method to bypass, or jailbreak, Fable 5’s safeguards.
The technique at the center of it is strikingly ordinary. It involved asking the model to read a specific codebase and identify software flaws, in other words, a code review. That same capability is something developers use every day, and Anthropic points out it is widely available from other commercial models, including GPT-5.5.
This is where the company and the government disagree on severity. Anthropic characterizes what was actually disclosed as “a small number of previously known, minor vulnerabilities,” says no disclosed jailbreak led to a harmful result, and frames the findings as minor rather than a novel weapon. The government treated the same information as serious enough to warrant pulling the model. The underlying technical details have not been independently verified.
Anthropic’s Defense of the Model
Anthropic’s statement reads less like an apology and more like a defense of the safety work behind Fable 5. The core argument is that these are the strongest safeguards it has ever shipped.
The company’s main claims:
- Its safeguards are “substantially more effective than those of any previously deployed model.”
- It used a defense in depth strategy, pairing narrow jailbreak resistance with active monitoring.
- An external bug bounty ran more than 1,000 hours and produced no universal jailbreaks.
- A mandatory 30 day data retention policy lets it study and shut down novel attacks, even for customers who normally hold zero retention agreements.
- The capability the government worried about already exists in other commercial models, so removing Fable 5 does not remove the capability from the world.
That last point is the crux of Anthropic’s position. The argument runs that a model this capable was going to exist regardless, and that shipping it with hard guardrails, fallback routing, and monitoring is safer than the alternative of a competitor releasing something equally strong with no protections at all. The government’s intervention is, in effect, a real world stress test of whether that argument holds.
A Quick Refresher on Fable 5 and Mythos 5
If you are coming to this story cold, here is what got pulled.
Claude Fable 5 launched on June 9, 2026 as the first publicly available model in a brand new top tier Anthropic calls the Mythos class, a fourth step that sits above Haiku, Sonnet, and Opus. It debuted at the top of independent leaderboards, posted state of the art results on most benchmarks it was tested on, and is built for long running autonomous work rather than quick chat. We covered the launch in full in our Fable 5 guide.
Mythos 5 is the same underlying system with the safety layer removed, available only to a small set of vetted partners such as cyber defenders and approved biology researchers. Fable 5 is what you get when Anthropic wraps that Mythos class capability in classifiers that block high risk requests in cybersecurity, biology, chemistry, and distillation, and quietly routes them to Opus 4.8 instead.
The pricing tells you where it sits. Fable 5 ran at $10 per million input tokens and $50 per million output tokens, exactly double Opus 4.8, and it was free to use on Pro, Max, Team, and Enterprise plans through June 22 as an introductory window. That window is now academic while the model is suspended.
Who Is Affected
Because the suspension is at the model level, the blast radius is wide and even.
- Every paid Claude subscriber who had Fable 5 in their plan
- API and Claude Platform developers calling the model directly
- AWS Bedrock customers in the US East and Europe regions where it had launched
- GitHub Copilot users who had access to it
- Trusted partners on Mythos 5, including the Project Glasswing cohort
What is not affected is just as important. Claude Opus 4.8, Sonnet 4.6, Haiku, and every other Anthropic model are running normally. If your workflow did not specifically depend on Fable 5 or Mythos 5, you will not notice anything at all.
When Will It Come Back
Anthropic is treating the suspension as temporary. The company says it is “working to restore access as soon as possible,” but it has not committed to a date, and the timeline is not entirely in its hands since the directive came from the government.
Reporting around the order suggests officials pointed to a window of “coming weeks” for additional security hardening before access could be restored. Until Anthropic or the government says otherwise, the responsible assumption is that Fable 5 and Mythos 5 are down indefinitely and you should plan around Opus 4.8 or another model for now.
Why This Matters Beyond One Error Message
A broken model selector is the surface of a much bigger story.
This is the first time export control machinery, the same toolset used to restrict advanced chips and sensitive technology, has been pointed at a generally available AI model and used to take it offline. It sets a precedent that a frontier model can be launched, gain real users, and then be pulled by government order in a matter of days.
It also exposes how concentrated the risk is. When a single directive can switch off the model you built your workflow around, for every user at once, with no warning and no committed return date, single vendor dependence stops being a convenience question and becomes an operational one. The teams least disrupted today are the ones that already had a second and third model they could fall back to.
That is the practical takeaway. Keep more than one frontier model within reach. Fello AI puts Claude, GPT, Gemini, DeepSeek, Perplexity, and more in one native app for Mac, iPhone, and iPad, so a suspension, an outage, or a rate limit on any one provider never leaves you stuck.
