“`html
Anthropic has deactivated its two most advanced AI models, Fable 5 and Mythos 5, following an export control order from the U.S. government issued late on June 12, which instructed the firm to restrict access for any foreign individual, whether within or beyond the United States, including Anthropic’s own foreign-national staff.
Since the organization claims it cannot effectively distinguish foreign users from the rest of its clientele in real-time, the actual outcome is a global shutdown of both models. All other Anthropic models continue to be accessible.
Access to Anthropic Fable 5 and Mythos 5 Blocked
The U.S. government, referencing national security officials, transmitted the order to Anthropic at 5:21 PM ET on Friday, June 12, 2026. The directive dictates that no foreign individual, whether inside or outside the United States, including Anthropic’s own foreign national staff, is allowed access to the two models.
Due to the impracticality of selectively enforcing nationality-based access controls on a large scale, Anthropic broadened the restriction to cover its entire global customer pool to ensure complete legal compliance. Access to all other Anthropic models remains unchanged.
The directive did not furnish specific information regarding its national security concern, but Anthropic’s interpretation is that the government believes it has discovered a method to bypass or “jailbreak” Fable 5.
The purported jailbreak strategy, as explained to Anthropic, fundamentally involves directing the model to analyze a specific codebase and pinpoint software vulnerabilities. According to Anthropic, the government has so far supplied only verbal indications of a potential narrow, non-universal jailbreak.
Anthropic examined a demonstration of the particular method and concluded that it revealed only a few previously recognized, minor vulnerabilities. Notably, the company asserts that other publicly available models, such as OpenAI’s GPT-5.5, can generate the same output without necessitating any bypass.
Anthropic has publicly conceded that achieving perfect jailbreak resistance is not feasible for any model provider at present. In response, the organization has adopted a defense-in-depth approach: engineering non-universal jailbreaks to be narrow in focus and universal jailbreaks to be prohibitively costly to create, paired with extensive monitoring to identify and terminate successful attacks swiftly.
As a component of this approach, Anthropic has established a 30-day customer data retention policy for Mythos-class models to assist ongoing jailbreak research and mitigation.
Anthropic has expressed regret to impacted customers and indicated it is working diligently to restore access, with a pledge to provide further technical details within 24 hours of the directive.
“`