US Shuts Down Anthropic's AI Models Over 'Jailbreak' Security Concerns

The US government has ordered Anthropic to suspend access to its advanced AI models Fable 5 and Mythos 5, citing national security concerns over a reported 'jailbreak'. The models' access was disabled for all customers to ensure compliance, despite Anthropic stating that the issue was minor and other models can perform similar actions without breaking anything.

The U.S. government has issued an export-control directive ordering Anthropic to suspend access to two of its advanced AI models, Fable 5 and Mythos 5, citing national-security concerns tied to a reported model "jailbreak" . Anthropic has disabled access to the affected models for all customers to comply with the order, even as the company characterized the underlying issue as minor and noted that other widely available models can perform similar actions .

The action is notable both for its scope and its precedent. Applying export-control machinery to a frontier model's availability, rather than to chips or hardware, signals a more assertive federal posture toward AI capabilities and adds a new layer of regulatory risk for model developers . It lands as Anthropic pursues a high-profile IPO and continues to expand commercial partnerships, making government posture a material variable for the company's trajectory.

For the broader AI sector, the episode sharpens the tension between rapid capability releases and security oversight. The items to watch are the duration and conditions of the suspension, whether similar directives extend to other frontier labs, and any downstream impact on enterprise customers that had integrated the affected models into production workflows.