In a significant move, the U.S. government has ordered Anthropic to disable its advanced AI models, Fable 5 and Mythos 5, for all users globally. This directive, issued in the late hours of June 12, 2026, stems from national security concerns, restricting access to any foreign national, including employees within the company.
Global Impact of the Directive
The directive’s implementation means that, due to operational limitations, Anthropic has extended the access block to its entire global user base. The company explained that real-time differentiation of foreign users from others is not feasible, prompting a blanket restriction. While Fable 5 and Mythos 5 are affected, Anthropic’s other AI models continue to operate without interruption.
Underlying Security Concerns
Although specifics of the security threat were not detailed, the government expressed concerns over possible vulnerabilities in Fable 5. The alleged issue involves a jailbreak technique that could exploit the model’s codebase to uncover software weaknesses. Anthropic has been informed verbally about these potential risks, which appear to be limited in scope and effectiveness.
Upon review, Anthropic found that the identified vulnerabilities were minor and previously known. They also noted that similar outputs could be generated by other AI models available in the public domain, such as OpenAI’s GPT-5.5, without any need for bypassing security protocols.
Anthropic’s Response and Future Outlook
Anthropic has acknowledged the challenges in achieving complete jailbreak immunity. In response, the company has employed a defense-in-depth approach, aiming to make universal jailbreaks costly and difficult, while narrowly restricting non-universal attempts. This strategy includes robust monitoring systems to quickly respond to any breaches.
To support ongoing research into mitigating these concerns, Anthropic has implemented a 30-day data retention policy for its Mythos-class models. The company is actively working to resolve the access issues and has promised to provide more technical details within 24 hours of the directive.
Customers affected by this shutdown have received apologies from Anthropic, which is prioritizing the restoration of service. The situation highlights the delicate balance between technological advancement and national security in the AI sector.
