In a dramatic escalation of federal oversight into artificial intelligence, the United States government has issued an emergency export control directive forcing AI safety provider Anthropic to abruptly suspend all access to its premium models, Fable 5 and Mythos 5.
The directive, received by Anthropic at 5:21 PM ET on Friday, bans access to the two models by any foreign national, both internationally and within the United States – including Anthropic's own foreign-national employees. Because the restriction applies to individuals inside and outside U.S. borders, Anthropic stated it had no choice but to completely disable the models for all global customers to ensure immediate compliance.
The Contested "Jailbreak"
According to Anthropic, the government’s letter did not outline specific national security threats but is believed to stem from an alleged "jailbreaking" method, a technique used to bypass an AI model's safety guardrails.
However, Anthropic has openly pushed back against the severity of the government's claims. In an official statement, the company revealed it had reviewed a demonstration of the specific bypass technique in question and found it only identified a "small number of previously known, minor vulnerabilities."
Anthropic noted that the uncovered flaws were relatively simple, adding that multiple other publicly available, competing AI models are capable of identifying the exact same vulnerabilities without requiring any specialized jailbreak.
What is Affected?
While Fable 5 and Mythos 5 remain completely dark, Anthropic confirmed that access to all of its other legacy and mid-tier models remains unaffected and online.
The abrupt shutdown marks one of the most aggressive interventions by U.S. export control authorities on a domestic AI firm to date, signaling a tightening noose around the deployment of advanced, frontier-class artificial intelligence systems. Anthropic has not yet specified a timeline for when, or if, the models will be cleared for redeployment.
