Anthropic said it disabled access to its Fable 5 and Mythos 5 AI models after a U.S. Commerce Department export-control directive barred the company from distributing the systems to foreign nationals, including non-citizen employees. Anthropic said it believed the directive was based on a narrow jailbreak concern and that weaker Claude models were unaffected. Separately, after backlash over how its safeguards handle requests, Anthropic announced it will make “flagged” downgrades visible to users. The change includes API responses that state a reason when requests are refused or downgraded for national-security reasons. Together, the developments highlight how compliance actions and model governance are shifting from back-end policy to user-facing product behavior—creating operational burdens for institutions and developers who rely on consistent AI access for research and instruction.