Deep State Kills Two AIs Overnight

The same federal government that cannot fix the border or balance a budget just reached into the cloud and switched off two of the most powerful new AI systems on earth in a single stroke.

Story Snapshot

The Trump administration ordered Anthropic to block foreign access to its new Fable 5 and Mythos 5 AI models, citing national security.
To obey the order, Anthropic disabled both models worldwide, even for American users and the U.S. government itself.
The government has not shared public evidence of a major threat, only hints of a narrow “jailbreak” tied to software hacking.
The move exposes a deep problem: unelected officials can now shut down key technologies with little transparency or accountability.

What The Government Just Did To Anthropic’s New AI Models

Anthropic launched its Claude Fable 5 and Claude Mythos 5 models as its most advanced “Mythos-class” artificial intelligence systems, built to run complex tasks and even work for long stretches on their own.^[3] Within days, the United States government sent an export control directive that banned access to these models by any foreign national, whether inside or outside the country, including Anthropic’s own foreign staff. Anthropic said the only way to fully obey was to turn both models off for everyone, worldwide.

Anthropic explained in its public statement that it is “complying with the government’s legal directive” and removing access to Fable 5 and Mythos 5 for all users, while stressing that other Claude models remain online. A detailed report notes the order came from the Commerce Department under national security powers and that it took effect almost immediately after Anthropic received it. Company leaders also said the government did not provide written technical proof of the risk behind the shutdown.^[1]

Why Officials Say This Is About National Security

According to coverage of the directive, the Commerce Department framed the move as an export control issue, much like past limits on advanced chips or encryption tools that could boost foreign militaries or hackers. Officials reportedly believe someone showed a way to “jailbreak” Fable 5 so it would more easily spot and fix software flaws in code, which in theory could help find security holes in critical systems.^[3] That kind of dual-use skill—helpful for defense, risky for offense—has often triggered heavy-handed controls in other tech fields.

At the same time, reporting based on Anthropic’s own description says the example the government pointed to was narrow, and used tricks other public models can already match, including those from rival labs.^[3] Anthropic has argued that no one has found a “universal jailbreak” that fully unlocks dangerous cyber powers in these models, and it views the model’s behavior as inside the normal bounds of what today’s top systems can do.^[3] That gap between a broad national security order and a narrow, disputed technical claim is what worries many observers across the political spectrum.^[1]

Anthropic’s Pushback And What It Reveals About Power

In its formal statement, Anthropic said clearly that it “disagrees” that the evidence it has seen justifies suspending both Fable 5 and Mythos 5 for every user. The company framed the shutdown as a compliance step, not an admission that its models are broken or out of control. It apologized to customers and promised to work to restore access “as soon as possible,” signaling that it sees this as a temporary, and perhaps unfair, interruption rather than a permanent recall. Tech commentators note that the firm even cut off access for the U.S. government to avoid violating the letter of the order.

This standoff shows how much power now sits with unelected officials inside agencies most voters never think about. A single export control letter can now switch off a key tool used by coders, small businesses, and researchers around the world. Many Americans on both the right and the left already feel the “deep state” protects its own power while regular families struggle. Actions like this, taken with limited public explanation, feed that distrust and raise hard questions about who really runs high-tech America.

Why Both Conservatives And Liberals Are Alarmed

For many conservatives, this looks like one more example of a federal bureaucracy that can micromanage private companies while ignoring core duties such as border control, crime, and debt. They see career officials deciding which tools Americans can use, even when the tools might help small firms compete with big tech giants. For many liberals, the move deepens fear that powerful technologies are being governed in closed rooms, without clear rules, public input, or strong protections for civil liberties and workers’ rights.

Three days.

That's how long Anthropic's Fable 5 was publicly available before the US government pulled the plug – globally, for every customer, including inside the US.

It launched June 9 as their most capable public model (Mythos class). Yesterday they received an… pic.twitter.com/S7tNDPwQBA

— Oscar (@OscarActual) June 13, 2026

Both sides can agree on one thing: almost no regular citizen had a say in whether these models should live or die. Anthropic says it had already built in strong safety rules and monitoring, including keeping data logs for its highest-risk systems to catch abuse. Yet the final call did not rest on a transparent, technical review but on a vague national security claim that the public still cannot see or test. That pattern, if it spreads, could shape not just artificial intelligence, but the balance of power between citizens and the state for years to come.