On June 12, 2026, Anthropic says it received a US government export-control directive and switched off two of its newest models, Claude Fable 5 and Mythos 5, for every user. The order restricts access by foreign nationals. The trigger, by Anthropic's account, was a narrow potential jailbreak that amounts to asking the model to read a codebase and fix its security flaws. Every other Claude model still works. This page sticks to Anthropic's official statement, because that is the only first-hand account of the order so far.
- Anthropic says it received the directive on June 12, 2026 at 5:21pm ET.
- It is described as an export-control directive citing national security authorities.
- The restriction applies to all foreign nationals, inside or outside the US, including Anthropic's own foreign-national employees.
- To comply, Anthropic removed access to Fable 5 and Mythos 5 for all users.
- Access to all other Anthropic models is not affected.
- The stated concern is a narrow, non-universal jailbreak: asking the model to read a codebase and fix its software flaws.
- Anthropic says the government provided only verbal evidence of it.
- Anthropic disagrees, says the capability is widely available from other models, and is working to restore access.
Two of Anthropic's newest models were suddenly switched off. If you tried to open Claude Fable 5 or Mythos 5 and found nothing there, this is why. I am writing it up from Anthropic's own statement and nothing else, because that statement is the only direct account of what the order actually said.
What actually happened
An export-control order, then a shutdown.
Anthropic says it received a directive from the US government on June 12, 2026 at 5:21pm ET. It describes the order as an export-control directive that cites national security authorities. The restriction is specific: it cuts off access to Fable 5 and Mythos 5 for all foreign nationals, whether they are inside or outside the United States, and that includes Anthropic's own foreign-national employees.
Anthropic says it complied by removing access to both models for every user. It also says access to all of its other models is not affected. So this is not a Claude outage. It is two specific models being switched off to satisfy a legal order, while the rest of the lineup keeps running.
The trigger: a coding prompt, not an exotic exploit
This is the part that should make developers look twice.
Here is what Anthropic says the government is worried about. The concern is a potential jailbreak, and by Anthropic's description it is narrow and non-universal. The method, in Anthropic's words, essentially consists of asking the model to read a specific codebase and fix any software flaws.
Read that again. The capability that pulled a frontier model off the market is, on its face, the thing a lot of us point these tools at every day: here is my code, find the weak spots, fix them. Anthropic also says the government provided only verbal evidence of this, through a shared report, rather than a documented, reproducible exploit.
Why this matters beyond the headline
If the line between a useful coding assistant and a national-security problem is "read this codebase and fix its flaws," that line sits right on top of a normal developer workflow. Whether you think the order is reasonable or an overreaction, it is worth noticing how ordinary the triggering behavior is.
Why everyone lost access, not just foreign nationals
The order targets a group. The fix targeted the product.
The directive restricts foreign nationals specifically. But Anthropic's response was to remove the models for all users. The statement frames this as what compliance required: to be sure it was not serving the restricted group, Anthropic switched the models off for everyone.
That is the practical reality of an access restriction drawn around who you are rather than what you do. A company cannot perfectly verify the nationality of every person behind every session in real time, so the safe way to comply with "no foreign nationals" is "no one, for now."
Not sure which AI model to use?
12 models · Personalized picks · 60 seconds
Anthropic's position: complying, but disagreeing
Following the order while pushing back on it.
Anthropic is clear that it is following the directive. It is equally clear that it thinks the directive is wrong. Its argument, from the statement, is twofold. First, the capability the government pointed to is widely available from other models, so singling out two Anthropic models does not remove it from the world. Second, Anthropic argues that applying this standard would essentially halt all new model deployments, because a narrow, non-universal jailbreak can be found in frontier models generally.
So Anthropic's framing is that this is a misunderstanding, not a genuine new danger, and it says it is working to restore access as soon as possible. It is complying because it is a legal order, not because it agrees.
What the statement does not answer
The honest gaps, kept honest.
I am not going to fill these gaps with speculation, because the only official account here is Anthropic's. A few things the statement does not settle:
- Which agency issued the order. Anthropic's statement does not name a specific official or department.
- The actual evidence. Anthropic says the government's evidence was verbal, so there is no public, documented exploit to evaluate.
- When access returns. The statement says Anthropic is working to restore access, but gives no date.
- The government's own account. There is no official government statement laying out its reasoning in public, at least not one Anthropic points to.
You will find plenty of reporting and online chatter that claims to know more. Some of it may turn out to be right. None of it is an official source, so it is not in this article.
What it means if you used Fable 5 or Mythos 5
Practical, not dramatic.
What is affected
| Status per Anthropic | |
|---|---|
| Claude Fable 5 | Suspended for all users |
| Claude Mythos 5 | Suspended for all users |
| All other Anthropic models | Not affected |
Source: Anthropic statement on the US government directive
If your work depended on Fable 5 or Mythos 5 specifically, you need a fallback for now, and the rest of the Claude lineup is still available. If you are choosing a coding setup more broadly, our Claude Code vs Cursor breakdown covers the current options, and the benchmark leaderboard tracks how the available models actually score.
This is also not the first time Anthropic has held a powerful model back. For the earlier chapter of this story, see our piece on Claude Mythos and Project Glasswing, and for the wider question of Anthropic and government pressure, our look at Anthropic, the Pentagon, and AI ethics.
The honest summary: a legal order took two new models offline over a coding-related concern, the company building them disagrees with the order and is trying to reverse it, and everything else still works. The rest is worth watching, but it is not settled yet.
Official source
One source, because that is the only first-hand one.
FAQ
Why were Claude Fable 5 and Mythos 5 suspended?
Anthropic says it received a US government export-control directive on June 12, 2026 at 5:21pm ET that restricts access to Fable 5 and Mythos 5 for all foreign nationals. To comply, Anthropic removed access to both models for every user. Anthropic says the government's concern is a narrow potential jailbreak that involves asking the model to read a codebase and fix its software flaws.
Are other Claude models affected?
No. Anthropic's statement says access to all other Anthropic models is not affected. Only Fable 5 and Mythos 5 were suspended. Claude Pro, Claude Code, and the rest of the lineup continue to work.
What is the Fable 5 jailbreak that worried the government?
According to Anthropic, the concern is a narrow, non-universal jailbreak that essentially consists of asking the model to read a specific codebase and fix any software flaws. Anthropic says the government gave only verbal evidence of it and that the same capability is widely available from other models.
Is Claude Fable 5 coming back?
Anthropic says it is complying with the directive but disagrees with it and is working to restore access as soon as possible. Its statement gives no restoration date, so the models remain suspended for now.
Keep Reading
Stay ahead of the AI curve
We test new AI tools every week and share honest results. Join our newsletter.