TL;DR: On June 12, the US Commerce Department ordered Anthropic to suspend foreign-national access to Fable 5 and Mythos 5. Anthropic complied by cutting off all global customers — a historic first for AI model-level export controls. Anthropic disputes the rationale, calling the evidence "verbal only" and the jailbreak concern "narrow and non-universal."

What Happened

On June 9, Anthropic released Claude Fable 5 to the public and launched Mythos 5 — the same underlying model with cybersecurity safeguards lifted — exclusively to vetted partners in Project Glasswing. The company described them as representing a new "Mythos-class" capability tier, the most powerful AI it had ever released.

Three days later, everything stopped.

On June 12 at 5:21 PM ET, US Commerce Secretary Howard Lutnik sent a letter to Anthropic CEO Dario Amodei announcing that the new models were now subject to export restrictions. Anthropic issued a statement shortly before 9 PM:

"The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees. The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance."

3 days From public launch to global shutdown
100% Customers affected (worldwide)
First ever US export controls applied to an AI model directly

The Government's Rationale: A Reported Jailbreak

The trigger appears to have been a report from another company claiming it had jailbroken Mythos 5. According to Axios, this alarmed US Commerce Department officials about potential national security risks — specifically the possibility that a successful jailbreak could give unauthorized users access to powerful automated vulnerability-discovery capabilities that Anthropic had reserved only for Project Glasswing partners.

💡
What is a jailbreak?
A technique to bypass an AI model's safety guardrails, coaxing it into performing actions it would normally refuse. Anthropic had built restrictions into Fable 5 for sensitive domains like cybersecurity and biochemistry, but the US government claimed these could be circumvented.

Anthropic Pushes Back Hard

Anthropic's statement disputed the government's characterization at every turn:

  • The government provided only verbal evidence of a "potential narrow, non-universal jailbreak."
  • The specific technique described involved asking the model to read a codebase and fix software flaws — capabilities already present in OpenAI's GPT-5.5 and other widely available models.
  • Those capabilities "are used every day by the defenders who keep systems safe."
  • Applying a "narrow potential jailbreak" standard broadly "would essentially halt all new model deployments for all frontier model providers."

The company said it believed there was a "misunderstanding" and it was working to restore access as soon as possible.

Why This Is Unprecedented

Factor Previous Export Controls This Action
Target Semiconductors (NVIDIA H100, AMD MI300X) AI software models themselves
Legal basis Computing capability restrictions National security, model-level capabilities
Scope Specific countries / end-uses All foreign nationals globally
Precedent Many prior instances First time ever for AI models

The US has long used export controls on AI-enabling hardware. The Commerce Department's Entity List has blocked chip sales to Chinese firms, and the Biden-era AI chip rules restricted advanced GPU exports. But controlling the software models themselves — the trained weights — is an entirely different regulatory frontier.

🌐
The AI Sovereignty Debate Ignites
The EU, which had just gained access to Mythos 5 after weeks of negotiations, reacted sharply. Thomas Regnier of the European Commission said the episode "further underlined Europe's need for technological sovereignty." The event demonstrated to every government worldwide how quickly access to frontier AI can be severed.

Broader Context: A Deteriorating Relationship

This didn't happen in isolation. Anthropic's conflict with the Trump administration dates to the company's refusal to allow its models to power fully autonomous weapons systems. In response, the Pentagon designated Anthropic a "supply chain risk" — the first time that designation, historically reserved for foreign adversaries, was applied to a US company. Anthropic sued the Pentagon, and a federal judge issued a temporary injunction blocking enforcement while litigation continues.

Key Takeaways
  • June 12: US Commerce Dept issues export controls on Fable 5 and Mythos 5 — historic first for AI model software
  • Anthropic immediately disabled both models for all global customers to comply
  • Government rationale: reported jailbreak poses national security risk
  • Anthropic disputes this: only verbal evidence, jailbreak is narrow, same capabilities exist in GPT-5.5
  • EU and global community: event proves need for "AI sovereignty" policies
  • Context: part of ongoing Anthropic vs. Trump administration conflict over autonomous weapons and the Pentagon blacklisting
🔗
Sources & Official References
— Anthropic Official: Claude Fable 5 and Mythos 5 — Launch Statement & Export Control Update
— BBC News: Anthropic suspends new AI tools over US government security concerns
— Nextgov/FCW: Anthropic Suspends Top AI Models After US Export Control Order
— Al Jazeera: US asks Anthropic to block global access to top AI models — why it matters