Anthropic has suspended access to Claude Fable 5 and Mythos 5 for all customers following US government security concerns raised just days after the powerful AI model's public release. The company confirmed it must "abruptly disable" both tools to ensure compliance with a federal order, marking an unprecedented escalation in Washington's scrutiny of leading AI developers.

The Jailbreak Claims

The Trump administration flagged what it described as a method of bypassing Fable 5's safeguards through "jailbreaking," the practice of circumventing software restrictions to access sensitive systems or unlock restricted features. Anthropic acknowledged the government's concerns but pushed back on the severity, stating that after reviewing a demonstration of the technique, they found only "a small number of previously known, minor vulnerabilities." The company emphasized that these flaws are "relatively simple" and that other publicly-available models can discover them without requiring any bypass.

A First-of-Its-Kind Designation

The suspension comes amid a broader clash between Anthropic and the Pentagon. US Defense Secretary Pete Hegseth labelled the company a "supply chain risk"—the first time a US firm has ever received that designation, historically reserved for foreign adversaries. The label effectively bars government agencies from deploying Anthropic's tools over security concerns. Anthropic is suing the Pentagon in response, and while a judge has ruled the directive cannot currently be enforced, the legal battle signals deep fractures between Washington and Silicon Valley.

International Fallout

The controversy has rippled across the Atlantic. The European Union, which gained early access to Mythos earlier this month following weeks of negotiations, said the episode underscores "Europe's need for technological sovereignty." Commission spokesman Thomas Regnier confirmed Brussels is "assessing" implications while the 27-nation bloc pushes forward with measures to reduce dependence on American and Asian AI providers. The irony isn't lost on observers: Anthropic built hype around Fable 5 by calling it "too powerful to release," yet now faces pressure from its own government for being too powerful to trust.

The Safety Paradox

Ahead of the public launch, Anthropic had marketed various safeguards designed to prevent cyber exploitation—warnings that critics dismissed as marketing spin. Gina Neff, Professor of Responsible AI at Queen Mary University London, noted the UK's AI Security Institute found the model could exploit defenses and systems 73% of the time during testing. "It's a step change in capability in cyber security," she told the BBC, adding that the industry has long warned these tools are advancing faster than safeguards can keep pace.

Key Takeaways

  • Anthropic disabled Fable 5 and Mythos 5 for all customers to comply with a federal order
  • The government flagged potential jailbreaking methods; Anthropic says vulnerabilities are minor and not unique
  • Pentagon labelled Anthropic a "supply chain risk" in an unprecedented move against a US company
  • A judge blocked enforcement of the directive while litigation continues
  • EU officials cited the incident as evidence Europe must reduce AI dependency on American firms

The Bottom Line

This is what happens when you build something too capable to ignore and too dangerous to trust—you get caught between government fear and market demand. Anthropic's self-proclaimed "too powerful" branding may have been clever marketing, but it's coming back to bite them in the form of regulatory backlash. The real question isn't whether Fable 5 can be jailbroken; it's whether anyone—including its creators—really understands what they've built.