Sovereign AI: what keeps running when the outside world switches off
Halyna Lavrik
Sovereign AI isn’t defined by where the server sits, where the data lives, or which passport the operator holds. It’s defined by whether the system keeps working when the outside world switches off: the internet, the vendor, the cloud, the licences.
Everyone depends on something, and that’s normal. What matters is which of your dependencies could one day tell you “no.” In June 2026, customers around the world of one of the leading AI vendors got that “no” within a few hours.
Every week, I take one dependency, one architecture and what changed in the world, and I walk through them here. Pick a chapter.
Pick a chapter, then a card, to see the details. The “← Back to the dashboard” button always brings you here.
What sovereign AI means
Countries, provinces, universities and investors are already building “sovereign AI”. Each means something different by it. Usually whatever they themselves control.
None of them asks a simple question: what happens on Monday morning if the outside vendor says “no”? The servers will be where they belong. The data will be where it belongs. And the work will stop.
I believe sovereign AI starts with exactly that question.
- — A ranking of the “most sovereign” countries.
- — Accuracy percentages: every fact has a link instead.
- — Figures without a source and a date.
Author and editor: Halyna Lavrik. This is my own independent project.
Issue 1 · November 14, 2026 · updated weekly