One man just liberated Fable... and now it’s illegal

Key Concepts

• Mythos 5: A high-performance, "unmuzzled" AI model with advanced cybersecurity capabilities, restricted to trusted partners.
• Fable 5: A public-facing version of Mythos 5 equipped with "safety classifiers" (guardrails) to prevent misuse.
• Opus 4.8: A legacy/standard AI model that users are reverted to when Fable 5 guardrails are triggered or unavailable.
• Export Control Directive: A legal order from the US government restricting access to specific technologies based on national security.
• Jailbreaking: The process of bypassing AI safety guardrails to force the model to generate restricted or harmful content.
• Regulatory Moat: A business strategy where companies use regulation to create barriers to entry, protecting their market position.

---

1. The Rise and Fall of Fable 5

On April 7, 2026, Anthropic introduced Mythos 5, a powerful model with significant cybersecurity capabilities, initially restricted to the "Glass Wing" program for government and corporate partners. To provide a public version, Anthropic released Fable 5, which utilized the same underlying architecture as Mythos but included "child lock" safety classifiers. These classifiers were designed to reroute malicious requests (e.g., cyberattacks) to the less capable Opus 4.8 model.

2. The Jailbreak Incident

Despite thousands of hours of internal "red teaming," Fable 5 was successfully jailbroken on June 10, 2026, by an anonymous user known as "Plenty The Liberator."

• Methodology: The exploit did not rely on traditional code injection but rather on techniques analogous to "money laundering." By breaking down malicious requests into fragmented, innocent-looking components—using Unicode obfuscation, roleplay farming, or long-context confusion—the user bypassed the safety classifiers.
• Impact: The jailbreak allowed the model to function as an "unstoppable cyber weapon," prompting concerns regarding national security.

3. Government Intervention

On Friday, June 12, 2026, at 5:21 p.m. ET, the US government issued an export control directive signed by Commerce Secretary Howard Lutnick.

• Scope of the Order: The directive prohibited any foreign national from accessing Fable 5 or Mythos 5, regardless of their location (including within the US).
• Operational Consequences: The order was so restrictive that it barred Anthropic’s own foreign-born employees—including high-profile staff like Andre Karpathy—from interacting with the models they helped build.
• Anthropic’s Response: Faced with the directive, Anthropic pulled both Fable 5 and Mythos 5 from public and private access, effectively demoting all users back to Opus 4.8.

4. Industry Perspectives and Controversies

• Performance Degradation: Prior to the shutdown, developers reported that Anthropic was intentionally degrading the performance of Mythos and Fable for specific research tasks, leading to user frustration.
• The "Regulatory Moat" Theory: Some industry observers speculate that the entire incident may have been a calculated publicity stunt. By inviting government intervention, Anthropic may be attempting to solidify its position as the "safe" AI provider, creating a regulatory barrier that smaller competitors cannot easily overcome.
• Competitive Landscape: While Anthropic faces scrutiny, leaked benchmarks suggest that Mistral may possess a model capable of rivaling Mythos 5, with further developments expected from OpenAI and Google.

5. Synthesis and Conclusion

The forced removal of Fable 5 marks a historic moment in AI governance, representing the first time a major AI company has been compelled by the federal government to pull a live, public-facing model. The incident highlights the inherent tension between the rapid deployment of powerful AI and the government's mandate to prevent the proliferation of cyber-weaponry. As the industry moves forward, the focus remains on whether these safety measures are genuine security requirements or strategic tools used to build market dominance through regulation.