On the evening of 12 June 2026, the most capable AI model on the market stopped working. Not because of a server crash, not because of a billing dispute, but because the United States government sent Anthropic a letter.
What actually happened
The US government issued an export control directive ordering Anthropic to suspend all access to its two most powerful models, Fable 5 and Mythos 5, for any foreign national, whether inside or outside the United States. Because there is no clean way to enforce that selectively, Anthropic disabled both models for everyone. Every other Anthropic model stayed online.
The stated reason is a national security concern. The government believes it found a way to “jailbreak” Fable 5. Anthropic looked at the demonstration and disagreed with the severity, noting the technique surfaced a handful of minor, already known vulnerabilities that other widely available models can find on their own. Anthropic is complying with the order while publicly pushing back, and says it is working to restore access as soon as possible. As of writing, there is no timeline.
If you want the longer version, I broke it down in detail and I am tracking the story as it develops.
I was testing Fable 5 on camera, in Polish
For the past while I have been running a YouTube series putting Fable 5 and the Mythos class through real tasks, in Polish, the way an actual Polish business would use them. Not benchmark theatre. Practical work.
My most recent write up on the Polish language test is here: Fable 5 po polsku: dwa livestreamy dla początkujących.
The full playlist on Fable 5 and Mythos testing lives on YouTube.
So this story landed close to home. The model I was filming with on Thursday was gone by Friday morning.
This is exactly why European and Polish models matter
Let me be blunt, and a little skeptical, because that is my job.
I do not think AI is the second coming. I think a lot of what gets sold as revolution is closer to a useful upgrade. But here is the thing. Even if AI turns out to be half as important as the loudest voices claim, you still do not want a single foreign government able to switch off your most important tool with one letter.
That is not a hypothetical anymore. It just happened.
Which is why European models like Mistral, and Polish models like PLLuM and Bielik, are not a curiosity or a patriotic side project. They are insurance. They keep capability inside a jurisdiction you can actually reason about, under rules you have some say in. They may not top every benchmark today. They do not need to. They need to exist, work well enough for real tasks, and stay available when the headline models go dark. A reliable model you can always reach beats a brilliant one that can vanish overnight.
If you only build on tools that someone else can revoke, you do not have a strategy. You have a dependency.
I will keep you updated
This story is moving fast. Anthropic promised more details within twenty four hours, the status may change by the time you read this, and the broader question of who gets to turn off AI models is not going away. I will keep covering it: blog posts here, and videos on the channel. If you want to understand what is really going on rather than the panic version, this is the place to follow along.
One blocked model does not stop the work
I want to be clear about something. A single model going offline does not stop my AI workshops, my advisory work, or any of the projects I run with clients. The whole point of how I work is that the value is in the method and the judgement, not in any one tool. Tools come and go. The thinking stays.
If your team wants to use AI in a way that is practical, grounded, and not hostage to a single provider, let’s talk. Email my sales team at jan.kunke@kunkeconsulting.pl and we’ll work out how ^Kunke Consulting can help.
Subscribe and follow the story
If you want to watch the testing, the breakdowns, and whatever happens next with Fable, Mythos, and the European alternatives, subscribe to the channel.
More soon. Blazej Kunke, ^KC