When a Government Pulls an AI Model: What the Fable 5 and Mythos 5 Suspension Means for Security Teams

On June 12, 2026, Anthropic disabled access to its newest models, Claude Fable 5 and Claude Mythos 5, due to a US government export-control directive. This action was taken to comply with national security authorities, citing a reported 'AI jailbreak' as the reason.

The directive ordered Anthropic to suspend all access to Fable 5 and Mythos 5 by any foreign national, regardless of their location. However, due to the difficulty in segmenting foreign nationals from US persons in real-time, the company opted for a blanket shutdown.

The 'jailbreak' in question involved asking the model to read a specific codebase and fix any software flaws. This capability is not unique to Fable 5 and is widely available in other models. In fact, defenders routinely use this capability for automated code review and vulnerability remediation.

The security community has emphasized that powerful capabilities like this are dual-use by nature. A port scanner or a packet analyzer can be used for both offensive and defensive purposes, depending on the context. The key is to manage these capabilities through coordinated disclosure, layered controls, and risk-based prioritization.

Anthropic's statement highlights the importance of transparency and fairness in government actions. The company agrees that governments should have the authority to block unsafe AI deployments, but only through a statutory process that is grounded in technical facts.

The suspension of Fable 5 and Mythos 5 has significant implications for security teams. It demonstrates the need for model redundancy and graceful fallbacks to ensure business continuity. Builders who had integrated Fable 5 into their workflows were affected by the sudden shutdown, emphasizing the importance of treating a single hosted model as a hard dependency.

To manage dual-use AI capability effectively, security teams should focus on securing the code produced by these models, putting guardrails around AI behavior, and treating AI systems as part of the overall attack surface. This includes scanning AI-generated code, implementing layered controls, and practicing coordinated disclosure.

Snyk's research has shown that nearly 4,000 AI agent skills carry at least one security flaw. This highlights the need for continuous monitoring and risk-based prioritization in managing AI-powered systems.

In conclusion, the suspension of Fable 5 and Mythos 5 serves as a reminder of the importance of model redundancy, coordinated disclosure, and risk-based prioritization in managing dual-use AI capability. By applying these principles, security teams can effectively manage the risks associated with powerful AI technologies and ensure business continuity.