Anthropic’s Newest AI Model Fable Just Shut Down Worldwide.
The US Government Just Shut Down Anthropic’s Newest AI Model Worldwide — Here’s What Actually Happened
Three days after launch, Claude Fable 5 and Mythos 5 went dark for every user on the planet — including people who had nothing to do with the reason given. This is the first time a government has used export controls to switch off a publicly deployed AI model, and the explanation is stranger than the headlines suggest.
On Tuesday, June 9, 2026, Anthropic released its most capable AI model ever — Claude Fable 5, the first public model in a new tier above Opus called the Mythos class. By Friday evening, three days later, it was gone. Not slowed down, not rate-limited — completely unavailable to every single user on the planet, after the United States government issued a directive that Anthropic says left it no other choice.
If you were using Claude around 10pm Eastern Time on June 12, you may have seen it happen in real time — the model simply stopped responding, replaced by an error or an “unavailable” message. That is exactly what occurred, and it is worth understanding properly, because the explanation is more unusual than “the AI was too dangerous.”
What Actually Happened — The Timeline
Anthropic publicly releases Claude Fable 5 — its first general release in the new “Mythos class,” a tier above Opus. Fable 5 went generally available in GitHub Copilot the same day and was described as state-of-the-art on nearly every benchmark.
According to Fortune’s reporting, Amazon’s security researchers reportedly identified that Fable 5 could be prompted, via a specific technique, to produce information potentially useful for cyberattacks. Amazon CEO Andy Jassy — Anthropic’s largest investor, a board member, and its cloud host — is reported to have relayed these findings to senior US officials.
Anthropic receives a letter from the US Commerce Department — an export control directive ordering the company to suspend all access to Fable 5 and Mythos 5 for “any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.”
Anthropic disables both models for every customer globally. AWS, which hosts Anthropic’s models, is asked to revoke access “for all users in all regions.” Other Claude models — including Opus 4.8, Sonnet, and Haiku — remain fully available and unaffected.
Anthropic publishes a detailed public statement disputing the government’s reasoning. As of writing, Fable 5 and Mythos 5 remain offline worldwide, with no confirmed timeline for restoration.
Watch — What Happened and Why
Why the US government shut down Claude Fable 5 and Mythos 5 — published June 2026. All rights respective owner. Source: YouTube.
Wait — Why Did “Block Foreign Nationals” Mean Blocking Everyone?
This is the detail that confused almost everyone, including AI policy experts who reacted with disbelief. The directive did not say “shut the model down.” It said: do not let any foreign national — anywhere in the world, including people physically inside the US, including Anthropic’s own non-American staff — access these two models.
Fable 5 and Mythos 5 are served through a shared cloud system — the same servers respond to requests from American and non-American users alike. Anthropic has no real-time way to verify a user’s citizenship before answering a prompt. Faced with a legal order to block “any foreign national” with no exceptions, and no technical mechanism to separate users by nationality on the fly, Anthropic’s only path to actual compliance was to turn the models off for absolutely everyone — American citizens included. The directive’s scope made a full global shutdown the only option, not a deliberate choice to punish all users.
That is the core of the story: an order ostensibly aimed at restricting foreign access ended up taking the product away from American users too, simply because there was no way to comply otherwise.
What Is “Mythos Class” and Why Does It Matter Here?
To understand why this specific model triggered this specific response, it helps to know what Fable 5 and Mythos 5 actually are. According to Anthropic’s own statement, Mythos is a new model tier that sits above Opus — previously Anthropic’s most powerful publicly available model. Fable 5 is Anthropic’s first general release built on this Mythos architecture, launched with additional safety restrictions. Mythos 5 itself — the more powerful, less-restricted version — remains limited to a small set of vetted research partners, not the general public.
Before this public launch, an earlier restricted research version called Claude Mythos Preview operated under something Anthropic calls Project Glasswing — used by a small number of trusted organisations specifically because of its advanced capabilities, including reportedly helping identify hundreds of software vulnerabilities for organisations like Mozilla during its restricted phase.
The capability at the centre of this controversy is exactly that vulnerability-finding ability — now built into a model available to “hundreds of millions of people,” in Anthropic’s own words, rather than a small vetted group.
So What Is the “Jailbreak” the Government Is Worried About?
This is where Anthropic’s account gets genuinely interesting — and where the company pushes back hardest. According to Anthropic’s official statement, the government has so far provided only verbal evidence of a “potential narrow, non-universal jailbreak.”
In plain English: a universal jailbreak would be a method that broadly defeats a model’s safety restrictions across many different requests — the kind of thing that would genuinely be alarming. A narrow, non-universal jailbreak affects only a specific, limited scenario. Anthropic describes the technique in question as, essentially, “asking the model to read a specific codebase and fix any software flaws” — something security professionals do every day as part of normal defensive work.
Anthropic’s central argument is that the capability identified is not unique to Fable 5 — the company says the same technique works on OpenAI’s GPT-5.5, which is not subject to equivalent export restrictions. If true, that raises an obvious question: why single out one model for a capability that other publicly available models reportedly also have?
The Bigger Picture — This Did Not Come Out of Nowhere
To understand why this escalated so quickly, it helps to know that Anthropic and parts of the Trump administration have been in an increasingly public dispute for months — reported by Fortune and other outlets as part of the broader context here.
In February 2026, President Trump reportedly ordered federal agencies to stop using Anthropic’s models after the company declined the Pentagon’s preferred contract terms — terms that would have allowed its models to be used “for any lawful purpose,” including autonomous weapons systems, which Anthropic had sought exemptions from. In early March, the Pentagon reportedly designated Anthropic a “supply chain risk,” a designation Anthropic is challenging in federal court. Several senior Trump technology policy figures have also publicly criticised Anthropic and its executives in the months prior to this directive.
Reaction from AI policy experts, as reported by Fortune, has been sharply divided. One former Trump administration AI policy advisor, Dean Ball, called the directive “simply cartoonish,” questioning how an administration that supports exporting advanced AI chips to China could simultaneously try to block the UK and “every other non-American on Earth” from using a US company’s AI model. Others took the opposite view — cybersecurity researcher Peter Girnus argued that Anthropic’s own marketing, which repeatedly described Mythos-class models in terms evoking weapons-grade capability, made this outcome predictable: “If you describe your product as a munition in every press release, eventually a government takes you at your word.”
What This Means If You Were Using Fable 5
According to Anthropic’s statement, only Fable 5 and Mythos 5 are affected. All other Claude models — including Claude Opus 4.8, Sonnet, and Haiku — remain fully available and were not part of this directive. If you were using Claude for general tasks via Sonnet or Opus, nothing has changed for you.
If you had integrated claude-fable-5 as a hardcoded model ID in an application — for example, via the API or a tool like GitHub Copilot, where Fable 5 went generally available the same day it launched — calls to that model ID will now return an error with no warning and no migration window. The practical fix reported by developers is the same one you’d apply to any unreliable dependency: treat the model ID as configuration rather than a hardcoded constant, so a single value change can redirect calls to an available model like claude-opus-4-8 or claude-sonnet-4-6 without a code change or redeploy.
What Happens Next
As of this writing, Fable 5 and Mythos 5 remain offline worldwide with no confirmed restoration date. Anthropic has stated it believes this is “a misunderstanding” and is working to restore access “as soon as possible,” and said it would share more details within 24 hours of the shutdown.
The timing is notable for another reason: Anthropic confidentially filed paperwork for a public stock market listing earlier this month, following a funding round that valued the company at $965 billion. Fortune reported that this export control action could make IPO investors more cautious — raising questions about whether a frontier AI company can be confident in its own product roadmap if the government can effectively recall a model worldwide with a single letter.
My Take — Mr Wangdoo
What strikes me most about this story is not the shutdown itself — it’s the mechanism. Governments restricting exports of sensitive technology is not new; export controls on encryption, military hardware, and dual-use technology have existed for decades. What’s genuinely new here is applying that mechanism to a cloud-hosted AI model in a way that, due to how cloud infrastructure works, ended up affecting the model’s home country’s own citizens as collateral.
Whether you think Anthropic’s safeguards were adequate or you think a company that marketed Mythos-class models around their cybersecurity capabilities should have expected this kind of scrutiny — both readings are represented in the expert reaction — the precedent itself is what matters going forward. This is now a documented case of a government using export control authority to take a publicly deployed AI product offline worldwide within hours of a directive being issued. Every frontier AI lab, and every business built on top of their models, now has to factor that possibility into their planning. That’s the part of this story that outlasts Fable 5 itself, whether or not it comes back.
Frequently Asked Questions
What is Claude Fable 5?
Claude Fable 5 is the first publicly released model in Anthropic’s new “Mythos class” — a tier positioned above Claude Opus, previously Anthropic’s most capable public model. It launched on June 9, 2026, was described as state-of-the-art on nearly every benchmark, and went generally available in GitHub Copilot the same day. It was built with additional safeguards compared to the more restricted Mythos 5, which remains limited to vetted research partners.
Why did the US government shut down Fable 5 and Mythos 5?
On June 12, 2026, the US Commerce Department issued an export control directive ordering Anthropic to suspend all access to Fable 5 and Mythos 5 for any foreign national, including foreign nationals inside the US and Anthropic’s own non-American employees. According to Anthropic, the government’s stated reason was a potential “jailbreak” — a way to bypass the model’s safeguards — related to its cybersecurity vulnerability-finding capability. Anthropic disputes that this finding justified the action and says it is a “narrow, non-universal” issue also present in other publicly available models.
Why did blocking foreign nationals take the model offline for everyone, including Americans?
Fable 5 and Mythos 5 run on shared cloud infrastructure that does not verify user citizenship in real time. The export control directive applied to “any foreign national, whether inside or outside the United States, including foreign national Anthropic employees” — with no carve-out mechanism. Because Anthropic had no way to selectively block only foreign nationals while keeping the service running for US citizens, the only way to achieve compliance was to disable both models for every user worldwide.
Are other Claude models affected?
No. According to Anthropic’s official statement, only Fable 5 and Mythos 5 are affected by this directive. Claude Opus 4.8, Claude Sonnet, Claude Haiku, and all other Anthropic models remain fully available and were not part of the export control order.
What is Mythos and Project Glasswing?
Mythos is the name for Anthropic’s most advanced model tier, positioned above Opus. Before its public release as Fable 5 and Mythos 5, an earlier restricted research version called Claude Mythos Preview operated under “Project Glasswing” — used by a small number of trusted organisations, reportedly including work that helped identify large numbers of software vulnerabilities for organisations such as Mozilla during its preview phase, due to its advanced cybersecurity-relevant capabilities.
Will Fable 5 come back?
As of this writing, there is no confirmed timeline. Anthropic has stated it believes the directive is based on a “misunderstanding” and is working to restore access “as soon as possible,” and said it would provide more details within 24 hours of the shutdown. Whether the models return depends on further discussions between Anthropic and the US government regarding the safeguards in question.
I built an app using the Fable 5 API — what should I do?
If your application has claude-fable-5 hardcoded as a model ID, calls to that model will currently return an error. The immediate fix is to point your application at an available Claude model — such as claude-opus-4-8 or claude-sonnet-4-6 — by changing a configuration value rather than your code structure, since the request and response format is consistent across Claude models. Going forward, storing model IDs as configuration rather than hardcoded constants protects against this kind of sudden unavailability regardless of the cause.
- Anthropic Official Newsroom — Statement on the US government directive to suspend access to Fable 5 and Mythos 5 (June 12, 2026)
- Fortune — Anthropic disables Fable and Mythos AI models after U.S. government bars it from giving foreigners access (June 13, 2026)
- CBC News — Anthropic says it has taken its latest AI models offline to comply with U.S. directive
- The Globe and Mail — Anthropic takes Fable 5 and Mythos 5 offline to comply with U.S. restrictions on foreign access
- CNBC — Anthropic disables access to Fable 5 and Mythos 5 to comply with government directive
- NBC News — Anthropic suspends new AI models after government directive
- Al Jazeera — US orders Anthropic to disable AI models for all foreign nationals
