On June 9, 2026, Anthropic released its most capable public AI model to date. By June 12, it was gone.
Not because it broke. Not because it underperformed. Because the US government issued an export control directive ordering Anthropic to suspend access for any foreign national, anywhere in the world, including inside Anthropic's own offices.
Three days of public availability. Then nothing.
Fable 5 was the first publicly available version of Anthropic's Mythos model, previously restricted to a small group of vetted partners due to its advanced capabilities.
What made it different was not just benchmark scores. It handled complex, long-running tasks that previous models could not sustain. It understood diagrams, charts, and tables embedded in files. It could implement designs with high fidelity, critique its own output visually, and update its approach based on what it learned during a task.
Developers who used it during those three days described it as feeling qualitatively different. Not just faster. More capable of doing actual work rather than responding to prompts.
It was priced at $10 per million input tokens and $50 per million output tokens. It came with a 1 million token context window and 128,000 maximum output tokens. It was available through the Claude API, on Amazon Bedrock, on Vertex AI, and on Microsoft Foundry.
Then it stopped being available anywhere.
At 5:21 PM ET, Anthropic received an export control directive from the US government citing national security authorities.
The order: suspend all access to Claude Fable 5 and Claude Mythos 5 for any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.
Anthropic could not verify citizenship in real time. The only way to comply was to shut both models down for every customer worldwide.
That is what they did.
Access to all other Anthropic models, including Opus 4.8, Sonnet 4.6, and Haiku, was not affected. But Fable 5 and Mythos 5 went offline globally with no published return date.
This is the first time in the history of commercial AI that a government export directive has pulled a specific model from the market days after general availability.
A jailbreaker operating under the name "Pliny the Liberator" posted on X on June 10 that he had bypassed Fable 5's safety classifiers using what he called a "pack hunt," a coordinated multi-agent attack.
The post went viral. The government directive appears to have followed shortly after.
Anthropic had already built hard limits into Fable 5 for high-risk areas including cybersecurity, biology, chemistry, and distillation. When those classifiers detected misuse, the model was designed to fall back to Opus 4.8. The argument from Anthropic was that the public version was meaningfully safer than Mythos 5, which ships without those guardrails.
The government directive did not distinguish between them.
The Fable 5 shutdown is not an edge case. It is an illustration of a structural risk that most product teams are not pricing in.
When you build on a specific model, you make a dependency decision. You tune your prompts around it. You test your outputs against it. Your users develop expectations based on it. Then something happens and the model is unavailable.
Sometimes that something is a capacity issue. Sometimes it is a pricing change. Sometimes it is a version deprecation. Now, apparently, sometimes it is a government order issued on a Friday afternoon with no warning and no timeline for resolution.
Every developer who built on Fable 5 between June 9 and June 12 was writing production code against a model that ceased to exist three days later.
The scale of this one is unusual. The underlying dynamic is not.
In the first quarter of 2026, more than 250 AI models were released in three months. Seven of them were considered top-tier at launch. Model quality, pricing, and availability are shifting continuously. The model that was the obvious choice in January is often no longer the obvious choice in April.
The organizations absorbing that volatility without disruption are not the ones that picked the right model. They are the ones that built so they do not depend on any single model to survive a disruption.
For our clients, the Fable 5 story is not primarily about Anthropic or US export law.
It is about one question: if the AI model or platform you depend on becomes unavailable tomorrow, how long does it take your business to recover?
If the answer is days, you have an infrastructure problem that has nothing to do with which model you chose.
The EU AI Act high-risk requirements taking effect on August 2 add another layer. Compliance documentation tied to a specific model version becomes immediately fragile the moment that version changes or disappears. The organizations that will handle that well are the ones treating AI infrastructure like infrastructure rather than like a vendor subscription.
Fable 5 will likely return in some form. The access suspension may be resolved. But the question of what happens to your operations in the gap is worth answering before the next one.
If you are a European SME thinking through AI implementation and you want to talk about how to build without creating single-model dependencies, that is the kind of conversation Itsavirus exists for.
.jpeg)
Claude Fable 5 was the first publicly available version of Anthropic's Mythos model, released on June 9, 2026. It was pulled three days later following a US government export control directive.
The US government issued a directive requiring Anthropic to suspend access for all foreign nationals. Because Anthropic could not verify citizenship in real time, the only option was to shut the model down globally.
Avoid building production systems that depend on a single model. Organisations that absorb disruptions well are those that treat AI infrastructure as infrastructure rather than a vendor subscription.
