Yesterday, on 12 June 2026, something happened that had never happened before – and it concerns everyone who works with artificial intelligence outside the United States.
At 5:21 pm Eastern Time, Anthropic, one of the world's leading AI companies, received a letter from the US government. The grounds: national security. The content: Fable 5 and Mythos 5 – currently the company's most capable models, a class above everything publicly available – may, with immediate effect, no longer be used by a single «foreign national». No matter whether in Zurich, in London or in Anthropic's own offices. Even the company's own foreign employees are locked out.
If you are not a US citizen, you are out. Full stop.
The practical result: because Anthropic cannot cleanly separate Americans from non-Americans in live operation, the company had to abruptly shut down both models for everyone. Hundreds of millions of users. Within hours. Worldwide.
This Is What Truly Alarms
A letter was enough.
No law. No procedure. No hearing. No verifiable technical evidence. The letter, according to Anthropic itself, contained no concrete details of the alleged threat. Verbally, there had been talk of a method that could «bypass» Fable 5's safeguards – a so-called jailbreak.
And what does this jailbreak consist of? You ask the model to read some program code and fix software bugs. In other words, exactly what security professionals around the world do every single day. Exactly what – according to Anthropic – OpenAI's GPT-5.5 and other freely available models can do just as easily, with no trick at all.
On this basis, the model that its own makers consider the safest ever released was taken offline for half the world.
What This Is Really About
Here, for the first time, something many had long only suspected becomes plainly visible: the US government has a direct line into the engine rooms of the American AI corporations. It can – overnight, without any public procedure, resting solely on the watchword «national security» – decide who has access to cutting-edge technology and who does not.
And we, outside the United States, belong to the second category whenever it suits them.
For anyone in Europe who relies on American AI – in the clinic, in the company, in public administration – this is a wake-up call. It is not about a single model. It is about the question of who holds the plug. Today it is Fable 5. What will it be tomorrow?
Anthropic is pushing back, calls the order a «misunderstanding» and wants to restore access as quickly as possible. The company rightly demands a procedure that is transparent, fair and grounded in technical facts – and notes that this order was precisely none of those things. Decent. But it changes nothing about the underlying finding: the switch sits in Washington. And it was flipped.
Digital Sovereignty Is Not a Slogan for Sunday Speeches
It is the very concrete question of whether an official letter from another continent is enough to cut us off from tools we use every day – without our being asked, without our being able to defend ourselves.
Yesterday at 5:21 pm, that question was posed for the first time in entirely practical terms. The answer should give us pause.