Government Shuts Down Powerful American AI

When the government can flip a switch and kill a leading American AI overnight, it confirms what many already fear: the system answers more to hidden security bureaucrats than to the people who will live with the fallout.

Story Snapshot

The U.S. government used a national security export rule to force Anthropic to shut down its most advanced AI models, Fable 5 and Mythos 5, worldwide, just days after launch.
Officials say a “jailbreak” could let the models help find and fix software flaws, while Anthropic argues the risk is narrow and similar to what other models already do.^[2]
To comply with a foreign-access ban that even covered its own immigrant staff, Anthropic disabled the models for all customers, including U.S. users and the U.S. government itself.
This fight highlights a deeper problem both left and right worry about: powerful agencies making sweeping tech decisions in secret, with little explanation, after Big Tech already rolled the product out.

What Exactly Did The Government Do To Anthropic’s New AI?

Anthropic, one of the top American artificial intelligence companies, launched two new models called Claude Fable 5 and Claude Mythos 5 in early June, calling Fable 5 its most capable model yet.^[7] Within about three days, the United States government sent an export control directive that blocked access to these models for any foreign national, whether inside or outside the country. Anthropic said the order even covered its own foreign-born employees. To be sure it followed the rule, the company disabled both models for every customer worldwide.

The directive came through national security authorities, not a public law passed by Congress, and the letter did not spell out detailed evidence of harm. Anthropic said its understanding is that officials believe someone found a way to “jailbreak” Fable 5 by having it read a code base and help find and fix software flaws.^[2] The company says it reviewed the example the government pointed to and concluded it showed only a small number of minor issues that other public models can also find without any special tricks.^[5]

Why Officials Say This Is About National Security

National security officials view cutting-edge artificial intelligence as a “dual-use” tool, meaning it can help defend systems but can also be misused to attack them. Reports say the Commerce Department used export control powers that usually apply to things like advanced chips, weapons technology, or strong encryption.^[1] The concern is that foreign adversaries could use an advanced model to discover software vulnerabilities faster and at scale, then turn those weaknesses into cyberattacks on American banks, power grids, or military systems.^[2]

Because the order banned use by all foreign nationals, not just people in specific countries, Anthropic said there was no practical way to separate “allowed” and “banned” users inside its own products and partner services. That technical reality, combined with strict penalties for export violations, led it to “abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.” Supporters of the directive argue that waiting for a proven breach could be too late, and that regulators sometimes must act on potential risk when systems are powerful and widely connected.^[1]

Why Anthropic And Many Users See Overreach And Double Standards

Anthropic did not present this shutdown as proof that Fable 5 and Mythos 5 were hopelessly unsafe. Instead, it said it was “complying with the government’s legal directive” while openly disagreeing with how broad the order was. The company described the jailbreak example as narrow and said the capability shown is already common across other publicly available models that remain online.^[5] It called the situation a “misunderstanding” and promised customers it was working to restore access as soon as possible.

Before the shutdown, Anthropic had already split the model into two versions to manage risk: Fable 5 for the public, with strong safety filters, and Mythos 5 with lifted cyber safeguards only for vetted defenders and critical infrastructure operators.^[7] In the public Fable 5, flagged cyber, biology, and chemistry prompts were routed down to a weaker, earlier model called Claude Opus 4.8.^[7] That design supports the company’s claim that it took safety seriously and tried to prevent casual misuse, raising questions about why these models were singled out while other tools with similar abilities stayed untouched.

What This Fight Reveals About Power, Elites, And A Broken System

This clash hits a nerve that many conservatives and liberals now share: a sense that unelected “experts” in Washington and wealthy tech firms make huge decisions together, then tell everyone else after the fact. The models were released to the public first, hyped as a major leap in capability, and only then did the government suddenly pull the plug under a secretive national security process.^[7] Small businesses, independent developers, and even some U.S. agencies using Fable 5 and Mythos 5 saw their tools vanish overnight, with no real say in the choice.

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

To many on the right, this looks like classic deep state behavior: powerful security bureaucrats using broad laws to control technology without clear oversight or limits. To many on the left, it feels like another case where big corporations pushed risky products out quickly, then expected the public to bear the cost when something went wrong. For both sides, the result is the same pattern they already distrust: decisions that shape the economy, jobs, and even national security are made in closed rooms, between elites, while ordinary Americans just live with the fallout.