Received June 12 at 5:21 PM ET, the directive prohibited any foreign national from accessing the models — including Anthropic's own foreign national employees. Because real-time nationality segregation isn't possible, Anthropic took both models offline worldwide. The stated trigger was a potential narrow jailbreak: asking the models to read code and identify software flaws. Fable 5 had launched three days earlier, on June 9, priced at $10/$50 per million input/output tokens. Anthropic complied and pushed back simultaneously. The company's public statement argued the jailbreak is widely replicable from other models, and that a recall standard this low would effectively halt all new frontier model deployments. The phrase that matters: Anthropic called the order inconsistent with "process transparency, fairness, and technical grounding" — which is not a denial of government authority, but a challenge to the evidentiary bar. This is a meaningful precedent regardless of how it resolves: the framework for suspending deployed AI models via export control now exists and has been used.