The U.S. government has ordered Anthropic to suspend access to its AI models, citing national security concerns over vulnerabilities.
Summary
- The U.S. government has ordered Anthropic to stop access to its AI models, Fable 5 and Mythos 5, due to national security concerns over potential vulnerabilities.
- Anthropic is complying with the order but disputes the government's claims, asserting that the vulnerabilities are not unique to its models.
- The company warns that this directive could hinder future AI model releases across the industry.
The U.S. government has issued an emergency export control order mandating Anthropic to cease access to its recently launched AI models, Claude Fable 5 and Claude Mythos 5. This decision, made just days after the models' release, is based on concerns regarding national security risks associated with a potential jailbreak vulnerability.
This order restricts access to the models for all foreign nationals, including Anthropic employees, leading the company to deactivate these models for its entire client base to comply with the directive.
Although the government did not specify the exact national security threats, it indicated that officials had identified a method to bypass the publicly accessible Fable 5 model. The Mythos 5 model, which is designed for cybersecurity applications, had already been limited to select partners.
Anthropic has raised questions about the validity of the government's findings, claiming that the vulnerabilities cited are relatively simple and can be observed in other publicly available models such as GPT-5.5. The firm noted that the government has only provided verbal evidence of a specific, narrow jailbreak method, which involves the model analyzing a particular codebase, a capability already present in competing AI systems.
While Anthropic is adhering to the government's order, it expressed concern that such a precedent could stifle the introduction of new AI models across the sector. "If this standard were to be adopted throughout the industry," the company remarked, "it would effectively halt all new model launches for frontier model developers."
Access to Anthropic's other AI models remains unaffected, and the company is working to restore access to the restricted models as soon as possible.
Discussions with individuals both inside and outside the government have highlighted the situation with Anthropic.
— David Sacks (@DavidSacks) June 13, 2026
David Sacks, co-chair of the President’s Council of Advisers on Science and Technology, commented on X that a trusted partner of both Anthropic and the government had reported a jailbreak of the Fable model. He noted that the administration had requested Anthropic CEO Dario Amodei to address the issue, but Amodei declined. Sacks criticized Anthropic for prioritizing the model's availability over safety, contradicting its public stance on AI safety and regulation.
Earlier this year, Anthropic faced tensions with the U.S. government over its refusal to sign an agreement permitting extensive domestic surveillance and fully autonomous weapons systems. Following a fallout, the Department of Defense labeled Anthropic a “supply chain risk,” a classification the company is contesting in court. However, recent reports suggest that tensions have eased, with the government showing interest in utilizing Claude Mythos and other models.
