Anthropic Suspends Claude Fable 5 After US Government Issues National Security Directive

Anthropic has suspended Claude Fable 5 - its most capable AI model to date - after the US government issued a national security directive ordering the company to cut off access for all foreign nationals. The order, received Thursday evening, was broad enough that Anthropic chose to disable the model for every customer worldwide rather than attempt selective enforcement. All other Claude models remain available.

What the Government Ordered

According to Anthropic, the directive arrived at 5:21 PM ET on Thursday with limited explanation. The government's stated concern is that a method of "jailbreaking" Claude Fable 5 had been identified - a technique that bypasses the safety restrictions built into the model, potentially enabling it to assist with tasks it was designed to refuse.

In the jailbreaking context, the government's evidence appears narrow. Anthropic said it reviewed a demonstration involving asking the model to read a specific codebase and identify software vulnerabilities. The company says the capability shown is not unique to Claude Fable 5: other publicly available models, including OpenAI's GPT-5.5, can perform the same task without any bypass required.

These vulnerabilities all appear relatively simple, and we have found that other publicly-available models are able to discover them as well without requiring a bypass.

Anthropic statement, June 12, 2026

Anthropic Is Complying - But Disagrees

The company confirmed it is following the legal order and has removed access to both Claude Fable 5 and Mythos 5. But Anthropic was direct in its objection, arguing that the government's standard - recalling a commercial model over a narrow, non-universal jailbreak - is one that no frontier AI company could meet.

If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers.

Anthropic statement, June 12, 2026

The Model Anthropic Called "Too Powerful to Release"

Claude Fable 5 launched publicly just days ago after a carefully managed pre-release in April that gave select organisations early access for security testing. Before launch, Anthropic itself described Fable 5 as "too powerful to release" - a phrase that drew both attention and skepticism, with critics questioning whether it was a genuine safety concern or a marketing narrative.

Ahead of the public launch, Anthropic worked with the US government, the UK AI Safety Institute, and multiple private red-teaming organisations to probe Fable 5's safeguards. That testing, the company said, showed no universal jailbreak - meaning no technique that could broadly bypass the model's restrictions across a wide range of tasks. Anthropic had also required a 30-day data retention policy for Fable customers, an unusual and commercially costly condition it put in place precisely so it could quickly detect and shut down any successful jailbreak attempts.

A Difficult Political Backdrop

The suspension did not happen in a vacuum. Anthropic is already involved in an active lawsuit with the Trump administration after US Defence Secretary Pete Hegseth labelled the company a "supply chain risk" - a designation historically reserved for foreign adversaries and applied to an American company for the first time. A US judge has since ruled that the Pentagon's directive cannot be enforced while the legal challenge continues, meaning government-adjacent organisations can still use Anthropic's tools for now.

The decision to invoke national security export controls to pull Claude Fable 5 represents a significant escalation of that tension. Whether it is a legally defensible use of those authorities - or an overreach - will likely be tested in the courts.

What Happens Next

Anthropic says it is working to restore access as quickly as possible and plans to share additional technical details within 24 hours of the suspension. The company has not indicated a timeline for resolution. In the meantime, users of Claude Fable 5 and Mythos 5 will find both models unavailable - with earlier Claude models unaffected.

The episode raises a broader question the AI industry has not yet resolved: who decides when a model is too dangerous to operate, what evidence is required to make that call, and what process protects companies - and the public - from directives that may not be grounded in technical reality. Anthropic closed its statement with a pointed demand for exactly that kind of framework.

We believe the government should have the ability to block unsafe deployments, as part of a statutory process that is transparent, fair, clear, and grounded in technical facts. This action does not adhere to those principles.

Anthropic statement, June 12, 2026

The US Department of Commerce has not publicly responded. Saganote will update this story as more details emerge.