Models off. No warning. No explanation.
At 5:21 PM ET on June 12, 2026, Anthropic received a directive from the US government citing national security authorities and ordering the company to suspend all Fable 5 and Mythos 5 access by any foreign national, whether inside or outside the United States, including Anthropic’s own foreign national employees. Anthropic disabled both models for all commercial customers immediately. Access to every other Anthropic model remains unaffected.
The directive letter did not provide specific details of the national security concern. The issuing agency wasn’t identified in Anthropic’s public statement. What was provided: an order, a timestamp, and a scope that reaches across borders.
According to Anthropic, the company understands the directive to be related to a jailbreak method for Fable 5. Anthropic stated that its review of the technique found only minor, previously known vulnerabilities, ones also discoverable in other publicly available models without requiring a bypass. Anthropic warned that applying this enforcement standard broadly would impede future frontier AI deployments. Those are Anthropic’s characterizations, not independently verified assessments. The government hasn’t publicly confirmed its rationale.
Independent evaluation of Fable 5’s capabilities by Epoch AI remains pending. All capability claims for Fable 5 are vendor-reported at this stage.
The practical scope is broad. Any commercial customer using Fable 5 or Mythos 5 lost access the same evening the directive arrived. Developers with integrations built on those models are operating without a timeline for restoration. Foreign national employees at Anthropic, wherever they’re based, are locked out. EU-based operators using these models through Anthropic’s API face the same suspension as US customers: the directive’s foreign national scope creates cross-border access implications that EU compliance teams will need to document and report against their own regulatory obligations.
The catch is that no compliance playbook existed for this scenario before June 12. Export control directives applied to AI models are new territory. Enterprise procurement teams, compliance officers, and developers who built production systems on Fable 5 or Mythos 5 are working through what “immediate enforcement” means for their risk documentation and vendor dependency assessments.
This is the first use of national security export control authority to force a frontier AI model offline without disclosed justification. The pattern it establishes matters beyond Anthropic. Any frontier model, from any lab, is now demonstrably subject to instant executive suspension on undisclosed grounds. That’s a material change to the risk calculus for every organization that treats frontier model access as an operational dependency.
The real question is whether the government identifies the issuing agency and the legal authority invoked, because the answer shapes how compliance teams can respond. If this is Export Administration Regulations authority, the documentation requirements differ from those triggered by NSPM-11 or Defense Production Act mechanisms. Until the authority is named, the compliance response is necessarily incomplete.
Don’t expect resolution in days. Enterprise teams should document the disruption, map their Fable 5 and Mythos 5 integration points, and flag this as a new category of vendor-dependency risk in their AI governance frameworks, one that existing vendor risk management processes weren’t designed to cover.