The Day the US Government Switched Off AI: Anthropic's Models Go Dark

In an unprecedented move that sent shockwaves through the global technology community, Anthropic was ordered by the U.S. government to disable its most advanced AI models for all foreign nationals — effectively shutting them down worldwide.

On Friday, June 12, 2026, Commerce Secretary Howard Lutnick sent a letter to Anthropic CEO Dario Amodei requiring the company to suspend export of its frontier models, Fable 5 and Mythos 5, to all destinations worldwide and to all foreign nationals, wherever located. The order carried the threat of "prompt criminal and civil penalties" for noncompliance.

Because there is no reliable way to verify citizenship on the internet, Anthropic had no choice but to disable both models globally. Even some of Anthropic's own foreign-born employees lost access. By 5:21 p.m., the models were dark.

The Trigger: A "Jailbreak" Over Code Analysis

According to Anthropic, the government shared what it described as a "potential narrow, non-universal jailbreak" — essentially asking the model to read a specific codebase and fix software flaws. This capability, which cybersecurity defenders use routinely, was deemed sensitive enough to warrant a full export-control directive.

The Wall Street Journal reports the directive emerged from conversations between Amazon CEO Andy Jassy and U.S. officials, including Treasury Secretary Scott Bessent, after Jassy reported a possible jailbreak vector. Amazon, which has a major investment in Anthropic, played a key role in surfacing the concern.

A Watershed Moment for AI Governance

This is the first time a U.S. government agency has directly ordered an AI company to restrict access to its most advanced models on the basis of national security concerns. The implications are staggering:

For developers worldwide, it means that access to frontier AI is no longer guaranteed — it can be revoked overnight by a foreign government. For enterprises, it raises the specter of operational disruption if their AI infrastructure depends on models that could be suddenly restricted. For the global research community, it threatens to fragment AI development along geopolitical lines.

The EU Reacts: Sovereignty or Fragmentation?

The European Commission moved quickly to frame the incident as proof that Europe needs technological autonomy. Thomas Regnier, a Commission spokesperson, said the body is "assessing its implications, including for users in the European Union."

The timing could not be more consequential. Just one day earlier, on June 14, the EU launched its European Technological Sovereignty Package — a sweeping set of measures aimed at sharply reducing reliance on U.S. and Chinese technology. The UK's AI minister, Kanishka Narayan, echoed the sentiment on X: "I care about sovereign AI because it now decides our security."

As Kate Hanaghan of TechMarketView put it: "The cost of dependency stays invisible until it's too late."

The Broader AI Landscape: A Week in Context

While the Anthropic story dominated headlines, other major developments unfolded simultaneously. Z.ai launched GLM-5.2 on June 13 with a usable 1-million-token context window — five times larger than its predecessor — enabling whole-repository code analysis in a single session. Meanwhile, OpenAI announced the OpenAI Partner Network on June 14, a $150 million program to train 300,000 certified consultants and accelerate enterprise AI adoption across systems integration, consulting, and data services.

These parallel developments paint a picture of an industry at an inflection point: models are growing more capable and more accessible, while governments are simultaneously tightening their grip on who can use them.

What This Means for the Future

The Anthropic directive marks a turning point in how AI is governed. We are moving from an era where frontier models were treated as commercial products to one where they are classified as strategic assets — on par with nuclear technology or advanced semiconductors.

This shift will have profound consequences:

• Geopolitical fragmentation of AI development, with different regions building separate model ecosystems
• Increased investment in sovereign AI, as governments and enterprises seek to reduce dependency on foreign providers
• A new arms race not just in model capability, but in export controls and access restrictions
• Ethical dilemmas around who gets to benefit from the most powerful tools humanity has ever created

Anthropic executives are scheduled to meet with Trump administration officials at the White House this week. Whether a resolution emerges or the models remain dark, one thing is clear: the era of unrestricted AI access is over. The question now is what comes next — and who gets to decide.