2 min read AI-generated

Fable 5 and Mythos 5 Shut Down: The US Government Steps In

Copy article as Markdown

The Trump administration forces Anthropic to disable its most powerful models worldwide via an export control directive — over an alleged jailbreak.

Featured image for "Fable 5 and Mythos 5 Shut Down: The US Government Steps In"

This is the kind of news you read twice: On Friday evening at 5:21 PM ET, the US government served Anthropic with an export control directive banning all foreign nationals from accessing Fable 5 and Mythos 5 — whether inside or outside the United States. Including Anthropic’s own non-citizen employees.

The result: Anthropic disabled both models for every customer worldwide. Not just foreign users — everyone. Because selective compliance wasn’t technically feasible.

What Happened

Commerce Secretary Howard Lutnick sent a letter to Dario Amodei placing Fable 5 and Mythos 5 under export controls. The reason: another company had shown the government a method to bypass Fable’s safety mechanisms — a so-called jailbreak.

Anthropic pushes back hard. The company reviewed the technique in question and says it only uncovers known, minor vulnerabilities that other freely available models can find too. No universal jailbreak, no new capabilities beyond what OpenAI’s GPT-5.5 can already do.

Anthropic’s Defense

In a detailed statement, Anthropic lays out its position:

  • Thousands of hours of red-teaming with the US government, UK AISI, and external organizations before launch
  • No tester has found a universal jailbreak
  • Anthropic follows a “defense in depth” strategy: jailbreaks should be either narrowly scoped or very expensive to produce
  • 30-day data retention enables rapid detection and mitigation of attacks

The key line: “If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers.”

Context: Anthropic vs. the Pentagon

There’s history here. Earlier this year, Anthropic refused to give the US military unrestricted access to Claude for surveillance and autonomous weapons systems. The government responded by placing Anthropic on a supply chain blacklist. This export control directive could be read as an escalation of that conflict.

What It Means

All other Anthropic models — Opus 4.8, Sonnet 4.6, Haiku 4.5 — remain available. Only the Mythos class is affected.

But the signal is concerning: a government can take a commercial AI model used by hundreds of millions of people offline within hours — based on an undisclosed jailbreak report, without a transparent process.

Anthropic is working to restore access and calls the situation a “misunderstanding.” Whether that’s accurate or whether larger political forces are at play remains to be seen.


Sources: