Related: Anthropic Launches Claude Tag, Putting an AI Teammate Inside Slack
Anthropic‘s two most capable Claude models, Fable 5 and Mythos 5, remain offline nearly two weeks after a United States government order forced the company to pull them, and the situation has now reached a new marker. The complimentary access window Anthropic offered when the models launched has closed, and as of June 23, 2026 the systems are still suspended with no restoration confirmed. What began as an abrupt takedown has hardened into an extended outage that leaves paying customers without two frontier models and the wider industry weighing a precedent it did not expect.
The sequence moved fast. Anthropic launched Fable 5 and Mythos 5 on June 9, positioning them as its strongest models and offering them at no extra cost to subscribers for a limited period. Three days later, on June 12, the US government issued an export-control directive citing national security authorities, and the order was sweeping. It required Anthropic to suspend all access to both models for any foreign national, whether inside or outside the United States, a category that included some of Anthropic’s own employees. The directive arrived in the evening, and by midnight the models were gone.
The trigger, according to the government, was a security concern rather than the models’ ordinary capabilities. Officials had become aware of a method for bypassing the safeguards on Fable 5, a so called jailbreak that could let a user push the model past its intended limits. The directive treated that vulnerability as a reason to cut off access entirely, and because Anthropic could not selectively block only foreign nationals while keeping the service running normally, complying in any narrow way was technically infeasible. The only way to meet the order was to disable both models for everyone, which is what the company did.
Anthropic has made clear it disagrees with the decision. The company argued that the finding of a single narrow jailbreak should not be grounds for recalling a commercial model already deployed to hundreds of millions of people, and warned that applying such a standard across the industry would effectively halt all new model releases from every frontier provider, since no large model ships without some vulnerability being found eventually. Anthropic described the order as resting on a misunderstanding and said it was working to restore access as soon as possible, but the models have stayed down, and the company has not been able to say when or whether they will return.
The closure of the free window adds a commercial sting to a regulatory problem. Anthropic had given subscribers on its Pro, Max, Team, and Enterprise plans access to Fable 5 at no additional charge for a set period after launch, and that 13-day window has now expired with the model still unavailable, meaning customers neither got the full benefit they were promised nor have a clear date for getting it back. For a company that competes on the strength of its top models, having them pulled by government order during the very window meant to showcase them is a difficult position to be in.
The episode lands at the intersection of two forces that have been building all year, the rapid advance of frontier AI and the government’s growing willingness to treat that capability as a matter of national security. Export controls have long governed advanced chips, but a directive that reaches into a deployed software model and pulls it offline for an entire class of users is a different kind of intervention, one that suggests models themselves are increasingly being regulated like strategic technology. Other frontier developers are watching closely, because the same logic could in principle apply to any of them.
What happens next will shape how the industry reads the precedent. If Anthropic restores Fable 5 and Mythos 5 quickly once the security concern is resolved, the outage may be remembered as a sharp but contained shock. If the suspension drags on, it will stand as a warning that a frontier model can be switched off by order, and that the companies building these systems do not have the final say over whether their most powerful products stay available. For now, two of the most capable models in the world remain dark, and the people who built them are still waiting, like everyone else, to learn when they come back.
Your First 10 AI Skills
10 practical AI skills, copy-paste prompts and a 7-day plan to start using AI with confidence.
Download the guide →