---
title: "The US government ordered Claude Fable 5 pulled"
description: "Three days after launch, Anthropic had to switch off Claude Fable 5 and Mythos 5 — on US government orders. What happened, the disputed flaw, and what comes next."
date: 2026-06-12
category: AI
tags: ["AI", "Anthropic", "Claude", "LLM", "Security"]
url: https://uper.pl/en/blog/fable-5-pulled-by-us-government/
---

# The US government ordered Claude Fable 5 pulled — the reason? A flaw its rivals can find too

Three days after Anthropic unveiled Fable 5 as the most powerful model it had ever released to the public, the company had to switch it off — for everyone, worldwide, on a Friday evening. Not because of an outage or a breach, but because the US government told it to.

The letter reached Anthropic on June 12 at 5:21 p.m. Eastern. It was an export control directive, citing national security, ordering the company to cut off access to Fable 5 and its non-public twin, Mythos 5, for any foreign national — anywhere in the world, including the company's own employees. A ban worded that broadly couldn't be honored halfway. To be certain no foreign national could touch the model, Anthropic simply turned it off for all. In Claude Code, trying to run the `claude-fable-5` model now returns a message saying the model "may not exist or you may not have access to it." The API throws the same error. The other models — Opus, Sonnet, Haiku — keep working.

## The disputed flaw: a "jailbreak" that was really a code review

So what can Fable 5 do that earned it the attention of the Commerce Department? This is where it gets interesting, because Anthropic itself says: not much. The directive, the company writes, came with no specific justification. From what Anthropic has been able to piece together, the government believes it found a way to get around the model's safeguards — and that this "jailbreak" was dangerous enough to warrant pulling the product. Except the demonstration the company reviewed amounted to asking the model to read a particular code repository and fix the flaws it found. In doing so it surfaced a handful of minor, already-known vulnerabilities — the kind, Anthropic stresses, that plenty of other publicly available models will turn up without any jailbreak at all. The company points directly at OpenAI's GPT-5.5 and adds that this is exactly the everyday work of the specialists who keep systems secure.

There's one more detail Anthropic underlines: the evidence it got from the government was purely verbal. Nor was this a universal jailbreak that would crack the model open across a whole range of dangerous uses — just a narrow trick that works under specific conditions. The company says it has not been shown a single case in which such a flaw led to real harm.

## Anthropic complies with the order — and tears its logic apart

That leaves Anthropic in an awkward spot: it is complying with a legal order while picking its logic apart. Recalling a model used by hundreds of millions of people over a narrow, potential flaw is, in the company's view, a precedent that — applied across the industry — would halt the deployment of any new model by every serious player. Anthropic has long argued that the government should be able to block unsafe deployments, but through a process that is transparent, fair, and grounded in technical fact. This time, it says, none of those conditions were met. The company has called the whole thing a misunderstanding and said it is working to restore access, with more details promised within a day.

It's worth remembering that Fable wasn't shipped quietly. Before launch, Anthropic spent thousands of hours red-teaming its safeguards — alongside the US government, the UK's AISI, and outside teams — and maintains they proved more effective than in any earlier model. Some users even complained the restrictions were too tight.

## Who issued the order

Where did the order actually come from? According to NBC News, citing an administration official, the letter was signed by Commerce Secretary Howard Lutnick and prepared by people at the Commerce Department's Bureau of Industry and Security (BIS). The department did not immediately respond to requests for comment, and why foreign nationals in particular should be the threat here remained unclear. As NBC notes, this is likely the first time a leading AI company has taken a live, public model offline because of federal intervention.

## Why Mythos 5 disappeared alongside Fable

To understand why both models were swept up together, you have to rewind a few days. Anthropic only showed off Fable 5 and Mythos 5 on Tuesday, billing them as the strongest systems in its history. Both grew from the same foundation — the earlier Mythos Preview — but went out into the world through different doors. Fable 5 reached a broad audience with heavy guardrails, especially around cybersecurity and biology. Mythos 5, stripped of those brakes, went only to a narrow group of trusted partners in cybersecurity and critical infrastructure. The same engine, two different handles — and a single directive that shut both at once. Mythos Preview had already earned some good press in a security-research program, Project Glasswing; Mozilla alone claimed it had patched hundreds of vulnerabilities with its help.

I covered the launch itself and both models in more detail in [Claude Fable 5 and Mythos 5 — Anthropic's most powerful model](/en/blog/claude-fable-5/).

## Background: months of friction with the Trump administration

This shutdown doesn't land in a vacuum, either. Anthropic's relationship with the Trump administration has been tense for months. In February, the president and Defense Secretary Pete Hegseth moved to push the company's products out of federal agencies after it began demanding tougher rules on how the Pentagon used its technology; Trump fired back sharply on Truth Social. Anthropic sued the administration over free speech, secured a favorable ruling from a federal court in California, and the case is still moving through a court in Washington. Until recently the temperature seemed to be dropping — the Financial Times reported that the NSA was using Mythos for offensive cyber operations — and on June 2 Trump signed an AI executive order directing agencies to harden their cyber defenses and to build a mechanism for early, voluntary government access to the most powerful models. Ten days later, that same government ordered the most powerful of those models switched off.

## What happens next

For now, new sessions across Claude's products fall back to the default model or to Opus 4.8, and every `claude-fable-5` call returns an error. Anthropic says it wants to bring access back as soon as the matter is cleared up. The story is still developing — the company has promised more to come.

## Sources

1. **Anthropic, official statement, June 12, 2026**
[https://www.anthropic.com/news/fable-mythos-access](https://www.anthropic.com/news/fable-mythos-access)
2. **NBC News (Jared Perlo), June 12, 2026**
[https://www.nbcnews.com/tech/tech-news/anthropic-suspends-new-ai-models-fable-mythos-government-directive-rcna349901](https://www.nbcnews.com/tech/tech-news/anthropic-suspends-new-ai-models-fable-mythos-government-directive-rcna349901)
3. **CNBC, June 12, 2026**
[https://www.cnbc.com/2026/06/12/anthropic-disables-access-to-fable-5-and-mythos-5-to-comply-with-government-directive.html](https://www.cnbc.com/2026/06/12/anthropic-disables-access-to-fable-5-and-mythos-5-to-comply-with-government-directive.html)
4. **Fortune, June 13, 2026**
[https://fortune.com/2026/06/13/anthropic-disables-fable-mythos-export-controls-national-security-threat/](https://fortune.com/2026/06/13/anthropic-disables-fable-mythos-export-controls-national-security-threat/)
5. **The New Stack**
[https://thenewstack.io/us-gov-orders-anthropic-to-pull-fable-5-and-mythos-5-three-days-after-launch/](https://thenewstack.io/us-gov-orders-anthropic-to-pull-fable-5-and-mythos-5-three-days-after-launch/)
6. **9to5Mac, June 12, 2026**
[https://9to5mac.com/2026/06/12/anthropic-pulls-claude-mythos-5-and-claude-fable-5-following-us-government-directive/](https://9to5mac.com/2026/06/12/anthropic-pulls-claude-mythos-5-and-claude-fable-5-following-us-government-directive/)
7. **Simon Willison**
[https://simonwillison.net/2026/Jun/13/us-government-directive-to-suspend-access/](https://simonwillison.net/2026/Jun/13/us-government-directive-to-suspend-access/)

*As of June 13, 2026.*