Anthropic will disable Fable 5 and Mythos 5 for everyone after export-control letter

Anthropic is about to pull the plug on its newest flagship AI models, and it is doing it because a US export-control order would bar foreign entities and individuals from using them. In a blog post on Friday evening, the company said the government’s letter arrived at around 5:21 p.m. ET and included “any foreign national inside or outside the US, including foreign national Anthropic employees.” Anthropic’s conclusion is blunt: the “net effect” of the order is to abruptly disable access to its models for everyone, not just for foreigners.

The models caught in the blast radius are Fable 5 and Mythos 5, released this month. Anthropic says it plans to disable access to those two models following the government order that would restrict foreign use due to national security concerns. The letter, the company added, “did not provide specific details” of the concern. Anthropic also disputed the issue’s severity, arguing the government’s worry appears tied to a potential way to “jailbreak” Fable 5.

To understand why this hits so hard, it helps to see what an export-control compliance problem looks like for a frontier model provider. Export controls are usually about who can use a system, where they are, and under what circumstances. But when a model is accessible through a single product surface, “disallow foreigners” can become a messy technical and operational burden. Anthropic appears to be taking the cleanest path for compliance: if the rules are triggered by nationality, it says it will disable access across the board for everyone using those models to ensure compliance.

Anthropic’s blog also matters because it frames the specific technical concern as narrower than the government implied, at least in the company’s view. The company said the technique appeared narrow, not universal, and involved known vulnerabilities that could be identified by other publicly available models. In other words, Anthropic is not claiming there is zero risk; it is arguing the risk is constrained and not an existential flaw that would justify shutting down access broadly.

The legal and regulatory context here is also getting more intense. The move is described as the latest escalation in Anthropic’s clash with the Trump administration over AI safety, national security, and “the extent of government control over frontier AI models.” Earlier this year, in February, the Pentagon moved to designate Anthropic as a supply chain risk after the startup sought limits on its AI model for certain defense applications. Anthropic sued the Defense Department over that designation, and the source notes that two lawsuits related to the government’s supply-chain risk label remain pending.

This is where the second-order impact shows up for boards and executives at AI companies, even those not named in any letter. When governments start linking national security concerns to export-control enforcement, the consequences are not limited to policy documents. They can directly change product availability, revenue timing, and how customers plan deployments. For enterprise buyers and integrators, model access is not a “nice-to-have feature”; it is often an input into workflows, customer promises, and downstream systems. If the models are disabled abruptly, the operational fallback plan suddenly becomes everyone’s problem.

Anthropic also said access to its other models will not be affected. That detail is important for decision-makers calculating the blast radius: this is not a total shutdown, and it is not an across-the-board disablement. It is targeted to the latest top models, Fable 5 and Mythos 5. The company also said it is complying with the order but disagreed with the government’s finding. Notably, an Anthropic spokesperson did not say when exactly the company would disable access.

For investors, founders, and operators watching the frontier AI market, this episode is a reminder that “AI safety” and “national security” are not abstract talking points. They are becoming enforcement hooks that can force product behavior in real time. And for competitors, the strategic stakes are immediate: regulatory friction can reshuffle who can deliver capabilities when, who retains enterprise trust, and who can keep models accessible to paying customers. In an industry where speed and access are competitive advantages, Anthropic’s decision to disable Fable 5 and Mythos 5 globally is a stark signal that compliance decisions can override product momentum overnight.