The World's Strongest AI Model Returns After the US Government Shut It Down: The Fable 5 Story

The World's Strongest AI Model Returns After the US Government Shut It Down

On June 9, 2026, Anthropic released its most powerful model ever: Claude Fable 5. Just three days later, the model vanished entirely by order of the US government. And after 18 days of back-and-forth between the company and Washington, it returned yesterday, July 1, to service once more. This is the story of what happened — and why it matters to you, even if you have never written a single line of code.

What Is Fable 5, and Why All the Noise?

Fable 5 is not a routine update. Anthropic described it as the most powerful model it has ever released to the public, with capabilities that exceed any model it had previously made generally available. By the numbers, it tops nearly every well-known performance benchmark and ranked first on FrontierBench, the advanced coding evaluation from Cognition. Its core strengths: software engineering, knowledge work, computer vision, scientific research — and above all the long-horizon tasks that intelligent agents (AI Agents) depend on.

The model belongs to a new class Anthropic calls "Mythos-class," and it has a more powerful, less-restricted twin named Mythos 5, available only to a small set of partners under a defensive-cybersecurity program called Project Glasswing. The two share the same underlying model, but Fable 5 shipped with strong safety controls to make it safe for general use.

The First Shot: An Order That Took the Model Offline Within Hours

On Friday, June 12, the US government imposed export controls on Fable 5 and Mythos 5. In practice, these controls required Anthropic to block access for any non-US person, inside or outside the United States. The problem was that the order took effect immediately, and the company had no reliable way to verify a user's nationality in real time — so the only available solution was the hardest one: suspending access to both models for every user in the world at once.

The Spark: A Report From Amazon Researchers

The direct cause was a report prepared by researchers at Amazon, who found a way to bypass Fable 5's safeguards (what is known as a jailbreak): getting it to identify software vulnerabilities, and in one case producing code demonstrating how a specific vulnerability could be exploited. In an already tense political climate around China's and Russia's access to advanced AI, Washington read this report as a national-security risk.

The Twist: Anthropic Responds With Numbers

Here the story takes on a second face. Anthropic ran its own tests and reached a striking conclusion: far less capable models can do the exact same thing. Models such as Claude Opus 4.8, GPT-5.5, and Kimi K2.7 were able to identify the same vulnerabilities, and every model tested reproduced the same demonstration. The company's takeaway: the reported technique exposed no unique cyber capability at all — it was a borderline case for the safety controls, involving routine defensive work. In other words, the risk was overstated.

How Did It Come Back?

Despite its objection, the company moved fast. It trained an updated safety classifier that blocks the reported behavior in over 99% of cases, and if a request to Fable 5 is blocked it is automatically routed to Opus 4.8. Then the resolution came step by step: on June 26 the government approved restoring Mythos 5 for some US organizations; on June 30 the export controls were formally lifted; and on July 1 Fable 5 returned to the world via the Claude platform, Claude.ai, Claude Code, and Claude Cowork.

For subscribers on Pro, Max, Team, and select Enterprise plans, Fable 5 is available within up to 50% of weekly usage limits through July 7, and via usage credits after that.

The Price: What Did Anthropic Give Up to Get It Back?

The return was not free. The company came out of the crisis with two commitments: first, a unified industry framework for rating the severity of model safety vulnerabilities, in cooperation with Amazon, Microsoft, Google, and the rest of the Glasswing partners; and second, deeper collaboration with the government spanning pre-release testing, information sharing, and joint research — meaning the state is now a party to the model's lifecycle before it reaches the market.

Why Does This Matter to You?

Beyond the technical detail, the story carries three messages for any business owner or technology builder in our region. First, advanced models have become "dual-use" technology subject to the logic of export control and national security, so who gets access to the AI frontier is now a geopolitical question. Second, identity and nationality verification is coming as a condition for accessing the strongest models, so plan for added friction rather than assuming permanent open availability. Third, relying on a single provider is a risk: when Fable 5 vanished overnight, everyone who had built on it needed an alternative within hours, so designing your systems to switch between models is no longer a luxury.

The Bottom Line

Eighteen days were enough to sum up the phase AI has entered: the world's strongest model can be launched, banned, and brought back — not by an engineering decision, but by a political one. Fable 5 returned stronger in governance and safety, but the bigger lesson is that the AI frontier no longer belongs to engineers alone; it is now a table that governments sit at too.

Sources: Anthropic's official announcements on the launch of Fable 5 and Mythos 5 and on redeploying Fable 5 (anthropic.com), plus coverage from NBC News, PCMag, and The Hacker News.