Two days ago we published a post about Claude Fable 5 — Anthropic’s most powerful model, free to use until June 22. Today, that post is history. The US government ordered the global shutdown of the model just 72 hours after its launch, on June 12, 2026, at 5:21 PM ET.
The contradiction we anticipated became bigger than we imagined.
What Exactly Happened
The US government issued an emergency export control directive ordering Anthropic to suspend access to Fable 5 and Mythos 5 for all foreign nationals — without exception, including foreign national Anthropic employees within US territory.
The practical problem: Anthropic cannot filter foreign nationals from its user base in real time. Result: global shutdown. Not a single user in the world can access Fable 5 or Mythos 5 for the duration of the suspension. All other Claude models are unaffected.
The Official Reason: A National Security Jailbreak
According to the directive, the government identified a method to «jailbreak» Fable 5 — specifically, asking the model to read a codebase and identify software vulnerabilities. The concern: foreign actors could exploit this capability to attack US critical infrastructure.
The directive offers no specific evidence — only verbal communication about a «narrow, non-universal potential jailbreak.»
Anthropic Disagrees — and Says So Publicly
In a public statement, Anthropic was unusually blunt: it is complying because the directive is a legal order, but it disagrees that a narrow, non-universal jailbreak is sufficient cause to recall a model deployed to hundreds of millions of people.
Anthropic’s technical argument is strong: the vulnerabilities exposed by this exploit are already achievable using other publicly available frontier models — including OpenAI’s GPT-5.5. If the problem is malicious code analysis, Fable 5 is not the only vector. Shutting it down doesn’t close the risk. It just moves it to a different provider.
The Pattern Nobody Wants to See
This is the first time the US government has ordered the withdrawal of a deployed commercial AI model, in real time, globally, 72 hours after launch. The precedent it sets is alarming:
- Access to AI models can be revoked without notice by government decree.
- A single directive can cut off service to hundreds of millions of users simultaneously.
- The justification can be verbal, with no public evidence or transparent appeal process.
- The provider cannot refuse, even if it disagrees.
For businesses in Latin America building critical workflows on Anthropic models: this should be a wake-up call about dependence on AI infrastructure controlled by a single government.
The Irony of All of This
Anthropic built its reputation warning about the dangers of AI. It created a safety board. It testified before Congress. It positioned Claude as the «responsible» model. And when it finally launched its most powerful model — the government it was warning about pulled the plug before anyone could determine whether the model was as dangerous as claimed.
Or more dangerous than Anthropic admitted.
The truth, for now, is that we don’t know. And that is exactly the problem.
What Comes Next
Anthropic says it is actively working to restore access. No date confirmed. The government has not specified what conditions must be met to lift the suspension.
If you are a company or professional who made decisions based on Fable 5’s availability — now is the time to audit your AI stack and ask yourself what would happen if the model you depend on disappeared overnight.
