On June 12, 2026, the US government ordered Anthropic to switch off its two most powerful models, Fable 5 and Mythos 5, for every user. Anthropic complied within hours. Hundreds of millions of people lost access to the AI they work in every day, and the stated reason was a single suspected jailbreak.
If yesterday's news was Anthropic hedging about the future, this is the future arriving early.
What did the government just do?
On June 12, 2026, the US government issued an export-control directive suspending all access to Anthropic's Fable 5 and Mythos 5, for every user, US citizens included. Anthropic complied and disabled both models, while saying access to its other models is unaffected. Per the company, the order followed a suspected method of jailbreaking Fable 5.
Two things make this larger than a routine outage.
The first is the reach. Fable 5 is the model behind Anthropic's consumer products, used by hundreds of millions of people. Mythos is the cyber model the company already kept locked up because it could break into systems on its own. One was the everyday tool. The other was the loaded weapon. Both went dark at once.
The second is the mechanism. This was not a bug or a billing dispute. It was an export-control order, the same category of rule that governs missile parts and encryption. The government reached into a live commercial product and turned it off.
Why did the government suspend Fable and Mythos?
According to Anthropic, the government believes it found a way to jailbreak Fable 5, meaning a method to get the model past its safety limits. Anthropic disagrees that this justified the action, and said so plainly while complying:
"We disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people."
The company also apologized to its customers for the disruption. So the picture is a standoff: the regulator saw enough risk to pull the most-used AI in the country off the shelf, and the maker thinks the response was far heavier than the threat. You do not have to pick a side to see the point. The decision was not Anthropic's to make. It was made for them, fast, and everyone downstream lived with it.
What changes when AI falls under export control?
Export controls exist for things a government treats as dangerous in the wrong hands: weapons, surveillance gear, certain code. Putting a frontier AI model in that bucket says the state now sees these systems the same way.
That is a turn worth sitting with. For three years the public argument about AI has been about jobs, accuracy, and marketing. The argument the government just had was about whether a model is a weapon. And it answered by switching two of them off.
This is the same regulatory state our AI Law Tracker follows day to day, and the pattern holds. There is still no single federal AI law, only a patchwork of state rules and one-off federal actions like this one. When that kind of system moves, it does not move with a plan. It moves with a hammer, after the fact, in a way nobody can predict from the outside.
What does this mean for a business that runs on AI?
This is where it reaches your desk. The AI you depend on is not under your control, and it never was.
A company with thousands of engineers and a trillion-dollar valuation could not keep its own flagship model online for a full day once the order came. If Anthropic cannot promise availability, neither can any tool you have wired into your work. The model you built a workflow around can be gone by dinner, for a reason that has nothing to do with you.
That does not mean stop using AI. It means stop being a hostage to any single piece of it.
The professionals and businesses that come through a week like this intact are the ones who own the system, not just the subscription. They know how their work flows, so a tool can be swapped without the work stopping. They keep a human in the loop, so judgment does not vanish when a model does. They treat the model as a part they can replace, not a foundation they cannot live without.
That is the difference between using AI and running it. It is also the spine of the way we build with people: a working AI you own, built around your own tasks, on a process that survives any single model. We call it SAM, a Strategic AI Manager, and where you stand on that climb is what The 7 Levels of AI Proficiency measures.
What should you do this week?
Three checks, none of which need a budget. The goal is to find out, before the next directive, where you are exposed.
Name your single points of failure.
List every tool your work cannot run without today. For each one, write down a working alternative you could move to this week. If the list is blank, that is the exposure.
Own the workflow, not the tool.
Write down how your most important task actually flows, step by step, in plain language. A documented process moves to a new model in an afternoon. A process that lives only inside one tool dies with it.
Keep a human on the decision.
Automate the work, keep a person on the part that judges the result. That is the one piece no directive, outage, or model swap can switch off.
None of this is about fear of AI. It is about not betting your livelihood on a tool a government can turn off at 5 p.m.
One more thing
Yesterday, Anthropic spent $150 million and still would not promise the benefits of AI reach ordinary people. Today, the government switched off the AI that hundreds of millions of those people use. Read together, the message is not subtle. The people building this and the people governing it are both improvising, in public, at high speed, and you are downstream of every move.
You cannot control any of that. You can control whether your work depends on a single tool staying online, or on a system and a skill that outlast any one model.
"The model you built your work around can be gone by dinner, for a reason that has nothing to do with you. Own the system, not the subscription."
by Harrison Painter
When the next model goes dark, will your work stop with it, or keep moving?
Sources
- An update on Fable and Mythos access. Anthropic. June 12, 2026.
- What Claude Corps Reveals About the AI Economy. LaunchReady.ai. June 12, 2026.
- AI Law Tracker. LaunchReady.ai.
Frequently Asked Questions
What happened to Fable 5 and Mythos 5?
On June 12, 2026, the US government issued an export-control directive suspending access to both models for all users. Anthropic complied and disabled them, while saying its other models stay available and disagreeing that the move was warranted.
Why did the government suspend Anthropic's models?
Per Anthropic, the government believed it had found a way to jailbreak Fable 5, meaning a method to bypass its safety limits. Anthropic argued a narrow jailbreak did not justify recalling a model used by hundreds of millions.
What is an export control on AI?
Export controls are rules that restrict access to items a government deems sensitive, like weapons or certain software. Applying them to a frontier AI model treats that model as a controlled technology rather than an ordinary product.
How should businesses protect themselves from AI tools going offline?
Avoid single points of failure: know a working alternative for every critical tool, document how your work flows so it can move between models, and keep a human in the loop so judgment does not depend on any one system.
Find your AI Proficiency level
The free 7 Levels of AI Proficiency assessment places you across seven stages of AI capability. Under ten minutes. Research-backed scoring.