When the Government Pulled the Plug on AI

On Friday evening June 12, the U.S. Department of Commerce ordered Anthropic to immediately disable its most powerful AI models. What happened next set a precedent that every developer, business owner, and technologist needs to understand.

What Happened

At 5:21 PM Eastern on Friday, June 12, 2026 — just three days after Anthropic launched Claude Fable 5, its most powerful publicly available model — Commerce Secretary Howard Lutnick sent a letter to Anthropic CEO Dario Amodei. The letter invoked national security export-control authority and ordered Anthropic to immediately suspend all foreign nationals' access to both Fable 5 and Mythos 5.

The scope was breathtaking. It did not just cover users in other countries. It covered foreign nationals inside the United States, including Anthropic's own non-citizen employees. Because Anthropic had no way to verify the citizenship of every user in real time, the company made the only call it could: it disabled both models for everyone, worldwide.

Just like that, hundreds of millions of people lost access to the two most capable AI models on the planet.

The Models in Question

To understand why the government acted, you need to understand what these models are.

Anthropic knew this was too powerful to release broadly. So they created Project Glasswing — a vetted program where roughly 50 trusted organizations, including Amazon, Apple, Google, Microsoft, and CrowdStrike, could access Mythos for defensive cybersecurity work under strict controls.

The government disagreed.

Why the Government Acted

The official justification was vague. The Commerce Department's letter "did not provide specific details." But through reporting from the Wall Street Journal, Semafor, and other outlets, a picture emerged:

White House AI adviser David Sacks gave a different account on X the following day. He said a trusted partner reported the jailbreak to the administration, that they asked Amodei to either fix it or pull the model, and that Amodei refused. Sacks said "the ball was in Anthropic's court."

Anthropic disputes this characterization. A source close to the company reportedly said they were never given technical details about the jailbreak and never refused to fix anything.

The Irony

Here is the part that should keep every AI company's CEO up at night.

Anthropic built its entire brand on safety. They are the company that says "we think these models could be dangerous and we take that seriously." When they previewed Mythos in April, they explicitly said it was too powerful for unrestricted public access. They restricted it to vetted partners. They built Fable 5 with guardrails specifically to address the risk.

And then the government used Anthropic's own safety rhetoric as justification for shutting them down.

As cybersecurity researcher Peter Girnus put it: describing your product as a munition could lead a government to treat it as one.

OpenAI's Sam Altman, never one to miss an opportunity, had already called Anthropic's Mythos handling "fear-based marketing" back in April. When the shutdown happened, the irony was not lost on the industry.

What This Means for Everyone

This is not just an AI industry story. This is a precedent that affects everyone who builds on, depends on, or uses cloud-based services.

The Bigger Picture

Days before the shutdown, Amodei published a policy essay calling for government authority to block or reverse frontier AI models that fail independent safety testing. He got what he asked for — just not the way he imagined.

The question nobody has answered yet: what happens next time? If a narrow jailbreak on a consumer AI model is enough to trigger a nationwide shutdown via export controls, where does that line get drawn for every other AI company? OpenAI's GPT-5.5 can reportedly produce similar outputs. Google's Gemini has similar capabilities. Are they next?

Anthropic says the capability in question is "widely available in other public models" and is "routinely used by cybersecurity professionals." If that is true, then this was not about security — it was about power. And that should concern everyone regardless of which AI company they prefer.

What I Am Doing About It

This is one of the reasons I build the way I build. Code sovereignty matters. When your entire business depends on a third-party service that can be turned off by a government letter on a Friday afternoon, you do not control your business.

I use Claude every day. It is an extraordinary tool, and I believe Anthropic is genuinely trying to build AI responsibly. But this event reinforces what I have always believed: own your infrastructure, own your code, own your data. Build systems where the intelligence is a tool you use, not a dependency you cannot survive without.

The models that got shut down — Fable 5 and Mythos 5 — are the tip of the spear. My day-to-day work uses Opus and Sonnet, which were unaffected. But the principle is the same. If you are building something important, build it so that no single point of failure, whether it is a company, a government, or a model version, can take you down.