Anthropic has cut off access to its two newest AI models, Claude Fable 5 and Mythos 5, after receiving a U.S. government directive tied to national security concerns. The decision landed just days after the models were introduced, turning an ordinary product rollout into an immediate access dispute.
The company said it chose to remove access for all customers, not only users outside the United States, in order to avoid any loophole in complying with the order. The directive, delivered on Friday night, June 13, reportedly required Anthropic to suspend access for “any foreign national,” both inside and outside the U.S., including foreign-national employees.
A narrow jailbreak claim sits at the center
According to Anthropic, its understanding is that the government believes it found a way to bypass or “jailbreak” Fable 5. Anthropic says its own review of the demonstration suggests the technique is narrow and not universal.
The company said the demonstrated behavior appeared to show the model identifying a small number of minor vulnerabilities that were already known. It added that the issue seemed relatively simple and that other public models could find the same weaknesses without any bypass method.
In a blog post, Anthropic said the government has so far provided only verbal evidence of the alleged narrow jailbreak. The example, as described by the company, amounted to asking the model to read a specific code base and then fix an existing software flaw.
That detail matters because the dispute does not appear to involve a mass data leak or a broad cyber incident described in detail. Instead, the concern seems to focus on whether the model can help surface and repair software weaknesses in ways regulators consider sensitive.
Safety controls were already part of the launch
Anthropic says Fable 5 was designed with strong safeguards intended to reduce the chance of misuse by malicious actors. The company argues that those controls should significantly limit harmful use.
The model was also introduced with additional protection layers related to biology, chemistry, and cybersecurity. Anthropic launched Claude Fable 5 on Tuesday, June 9, as an update to the company’s Mythos AI model.
Before the full launch, the model had been available in limited preview in April. Anthropic also said its development was carried out in collaboration with the U.S. government, with the broader goal of easing concerns that the company’s AI technology could be exploited by bad actors.
A conflict that reaches beyond one model
The removal of Fable 5 and Mythos 5 is the latest flashpoint in a wider conflict between Anthropic and the Trump administration. Earlier this year, the Department of Defense under Trump designated Anthropic as a “supply chain risk” after the company restricted military use of its technology on ethical grounds.
That designation later barred U.S. government agencies from using Anthropic technology. The company then sued the U.S. government, adding a legal dimension to the long-running dispute.
Against that backdrop, the latest order is less a standalone technical issue than a collision of export controls, national security policy, and the government’s relationship with AI developers. For now, there is no public timeline for when Claude Fable 5 and Mythos 5 might return.
Anthropic has not announced when access will be restored, or what exact technical changes may be needed. Any return would likely depend on how the company addresses the vulnerability claim that the government says it identified.
