Three days. That's all it took for the US government to pull the plug on what Anthropic called its most capable AI models ever. Claude Fable 5 and Mythos 5 launched on June 9, 2026 — and by June 12, they were gone. Globally. For everyone. And the story behind it is messier, more political, and more consequential than most people realise.

What Launched — and What Got Shut Down

Fable 5 was Anthropic's first public release from its Mythos model family — a line the company had previously described as too powerful in the cybersecurity domain to release publicly. Mythos 5, the less restricted version, was available only to a vetted group of organisations through Anthropic's Project Glasswing programme.

Both models were positioned as a significant leap forward. Anthropic spent weeks red-teaming them with the US government, UK AISI, and multiple third-party organisations before launch. They believed the safeguards were the strongest ever deployed on a commercial AI model.

The Shutdown: What Actually Happened

On the evening of June 12, US Commerce Secretary Howard Lutnick sent a letter to Anthropic CEO Dario Amodei directing the company to suspend all access to Fable 5 and Mythos 5 for any foreign national — anywhere in the world, including foreign nationals working inside the United States.

The net effect: Anthropic had to disable both models for all customers globally to ensure compliance. The directive cited national security authorities and export control powers.

The Amazon Angle Nobody's Talking About

Here's where it gets complicated. Reporting from Semafor and the Wall Street Journal points to Amazon as the "trusted partner" who flagged the jailbreak to the government. Amazon CEO Andy Jassy reportedly told Treasury Secretary Scott Bessent and other officials that Amazon researchers used Claude Fable 5 to obtain information that could be used in cyberattacks.

Amazon is one of Anthropic's largest investors and provides much of its cloud infrastructure. The company that funds Anthropic — and profits from its success — reportedly helped trigger the shutdown of its most important new product. That's a remarkable dynamic, and one the industry hasn't fully processed yet.

Anthropic's Response: We Disagree

Anthropic pushed back hard. In its official statement, the company said it had reviewed what it believes to be the underlying report and concluded that "the level of capability demonstrated is available from other publicly deployed models, including OpenAI's GPT-5.5, and is used by cybersecurity defenders routinely."

White House AI adviser David Sacks told a different story, claiming Anthropic refused to fix the jailbreak when asked. Anthropic disputes this account entirely, saying it was never presented with specific details and never refused to fix any issue.

The Loud Take: This Changes Everything

Here's what matters beyond the specifics of this particular dispute: a sitting government has now demonstrated that it can switch off a widely used AI product, mid-deployment, on the basis of a security assessment that the company itself says is inaccurate — and with no transparent process, no advance notice, and no independent review.

Anthropic itself has publicly argued that governments should have the ability to block unsafe deployments — but as part of a statutory process that is transparent, fair, clear, and grounded in technical facts. This action, the company says, does not adhere to those principles.

There's also a notable political backdrop. Anthropic and the Trump administration have been in conflict since early 2025. The Department of Defense previously designated Anthropic a supply chain risk. And just one day before the shutdown, Anthropic CEO Dario Amodei published a major policy essay calling on the US government to hold legal authority to block or reverse frontier AI models that fail independent safety testing. Two days later, the government used that authority against Anthropic.

What Happens Next

Anthropic says it is working to restore access as soon as possible and believes the shutdown is a misunderstanding. The company is also navigating a commercially sensitive moment: it filed a confidential IPO prospectus with the SEC in early June, disclosing a revenue run rate of $47 billion and a valuation of $965 billion.

The broader question — who gets to decide when an AI model is too dangerous to deploy, and through what process — is now live in a way it wasn't before June 12. That question doesn't go away when Fable 5 comes back online.

Key Takeaways

• Anthropic launched Claude Fable 5 and Mythos 5 on June 9, 2026 — their most capable models ever, from the Mythos family previously considered too powerful to release publicly.
• Three days later, the US Commerce Department ordered both models taken offline globally, citing a reported jailbreak and national security concerns.
• Amazon — Anthropic's largest investor — reportedly tipped off the government, creating a striking conflict of interest.
• Anthropic disputes both the severity of the jailbreak and the characterisation of its response, arguing the same capabilities exist in other publicly deployed models.
• The shutdown sets a significant precedent: governments can now demonstrably switch off commercially deployed AI products mid-deployment, without transparent process or advance notice.
• All other Anthropic models (Opus, Sonnet, Haiku) remain available. Anthropic says it is working to restore Fable 5 and Mythos 5 access.