Anthropic Fable 5 — The Most Powerful AI Model Yet, Shut Down by US Government After Three Days

Launched to Great Acclaim

On 9 June 2026, Anthropic officially released Claude Fable 5 – the first model in the Mythos line to be made widely available to the public. It marked the first time ordinary users could access an AI series that had previously shaken the cybersecurity world due to its exceptional ability to find and exploit software vulnerabilities.

Fable derives from the Latin fabula – "a story told" – akin to the Greek mythos. The difference between Fable and Mythos lies in the protective layers:

• Fable 5: equipped with safety classifiers
• Mythos 5: the same foundation but without these filters, reserved for approved partners

In high-risk fields such as cybersecurity, biology, chemistry and nuclear science, Fable 5 automatically blocks responses and redirects to Claude Opus 4.8. Anthropic describes this as an unprecedented multi-layered defence in the history of their AI development.

Claude subscribers can try Fable 5 for free until 22 June without using credits.

Origins in the Secret Glasswing Project

Earlier, in early April, Anthropic had launched Mythos through Project Glasswing – a controlled-access sharing programme with strategic partners such as Apple and NVIDIA, aimed at helping these organisations strengthen software security against AI-driven threats.

Anthropic stated that the broad release of Fable 5 was made possible by new safeguards that block responses in high-risk areas. This also marked progress toward the "ultimate goal" the company had announced: deploying Mythos-class models on a larger scale.

The Shock: US Government Orders Shutdown

The celebration was short-lived as a storm hit.

At 5:21 pm on 12 June 2026 (ET), Anthropic received an export control order from the US government, requiring the suspension of all access to Fable 5 and Mythos 5 for customers worldwide. No warning. No transition period. Simply a shutdown.

This is widely recorded as the first time in history a leading AI company has had to take down an active model under a US federal directive.

The US government cited concerns over a jailbreak method that could bypass Fable 5's protections. Anthropic directly rebutted: the vulnerability cited was narrow, not universal, and the results obtained could already be achieved using other publicly available models such as GPT-5.5.

Unable to filter access by nationality in real time across all customer contracts and cloud pathways, Anthropic chose to suspend both models for all customers everywhere to ensure compliance.

Aftermath and Reactions

Anthropic's status page recorded an error at 00:50 UTC on 13 June, affecting:

• claude.ai
• Claude API
• Claude Code
• Claude Cowork

All other Claude models, including Opus 4.8, continued to operate normally.

Anthropic called this a "misunderstanding", stating it was working to restore access as soon as possible and had not committed to a specific date.

The developer community reacted strongly on social media. One developer wrote:

"My Fable, three days together were truly magical. Some things are so good the government has to step in."

A Major Lesson for the AI World

This incident raises serious questions about technological sovereignty and the risks of relying on a single AI model. For any organisation integrating AI into critical processes, this is the clearest demonstration in 2026 that a model can disappear overnight for reasons entirely beyond their control. Anthropic is still in negotiations with the US government. Fable 5 — the most powerful model ever publicly released — remains dormant in the shadows.

Compiled from NBC News, CNBC, TechCrunch, Engadget and official Anthropic documents.