Digital sovereignty is not about owning everything, it is about ensuring that nothing essential can be taken away.
That single line settles more arguments than it first appears to, because it disqualifies the two positions that dominate the debate. One side imagines sovereignty as control. The other imagines it as choice. Both are talking about something real, and both have mistaken a part for the whole.
The control camp misreads the goal
The instinct to equate sovereignty with control is understandable. Sovereignty, in its older political sense, was control: borders, currency, the monopoly on force. So when the word migrates into the digital domain, people reach for the same picture: own the servers, hold the data, build the stack, depend on no one.
The trouble is that this picture is both impossible and beside the point.
It is impossible because no institution, and certainly no continent, is going to reimplement the entire computing stack from silicon upward. The dream of total self-sufficiency is not sovereignty, it is autarky, and autarky has failed everywhere it has been seriously attempted.
A Europe that tried to own everything would spend a generation rebuilding inferior versions of things that already work, and would be no freer at the end of it.
But the deeper problem is that control is the wrong target even when it is achievable. An institution can own the building and still be a tenant of the lock.
A public administration can run its own data centre, on its own soil, under its own staff, and still find that every document it produces is hostage to a format only one vendor fully understands. The hardware is sovereign. The institution is not. Control over the container tells you nothing about who controls the contents.
This is why the control framing quietly concedes the argument before it begins. It accepts that sovereignty is about what you hold, and so it can always be answered with “but you cannot hold all of it”, which is true, and which is why the framing loses. Sovereignty was never about the holding.
The choice camp misreads the threat
The opposite error is more fashionable and, for that reason, more dangerous.
Here the argument runs: sovereignty simply means freedom to choose. A sovereign buyer surveys the market and selects the best tool for the job, and if the best tool happens to be proprietary, so be it. To exclude proprietary software on principle, this camp says, is itself a kind of unfreedom, an ideology dressed up as independence. Real sovereignty, they insist, is vendor-neutral.
It is a seductive argument because it borrows the language of liberty. But it confuses a choice made today with a choice that remains available tomorrow, and that confusion is the whole game.
Consider what “choosing” a proprietary, closed-format platform actually buys you.
On the day of purchase, you have exercised your freedom: you compared options and picked one. But every document, every workflow, every trained habit that follows accumulates inside a system that only its owner can reproduce.
Five years later, when the licence terms change, or the price rises sharply, or a feature you depend on is discontinued, or the company is acquired and the product withdrawn, you discover that your freedom to choose has quietly expired. The market has not removed your options on paper. It has removed the ground beneath all but one of them.
This is the trap the choice framing cannot see, because it measures freedom at the moment of decision and never at the moment of exit. A choice you cannot reverse is not really a choice, it is a commitment wearing a choice’s clothes. Sovereignty has always been about the capacity to reverse, to leave, to change one’s mind, to refuse a deal that has turned bad. A buyer who cannot say no later was never sovereign to begin with.
What is actually essential
So if sovereignty is neither owning everything nor choosing freely among everything, what is left?
What is left is the thing the opening sentence points to: ensuring that nothing essential can be taken away.
The word doing the work there is essential. An institution does not need to own the cloud, it needs to be sure that if a provider vanishes tomorrow, its data does not vanish with it.
It does not need to forbid proprietary tools it needs to be sure that the things it commits to those tools – its records, its contracts, its citizens’ files – remain readable, movable, and re-creatable by someone other than the vendor who sold it the tool.
Sovereignty is not a wall around one’s possessions. It is a guarantee about one’s exits.
This is why open standards, and not ownership, are the real substance of digital sovereignty, and why open standards alone are not even enough.
A format must do two things. It must persist: it must be readable in ten years without asking permission. And it must be reimplementable: anyone, in principle, must be able to build a tool that reads and writes it, without reverse-engineering, without a licence, without the original vendor’s blessing.
Persistence without reimplementability is a museum exhibit: you can look at your data, but only one company can do anything with it. Reimplementability is what turns a surviving file into a living, portable, sovereign asset.
The shape of the mistake
The control camp and the choice camp look like opposites. One wants to build a fortress, the other wants to shop in an open market. But they make the same underlying error: they locate sovereignty in the present tense: in what you hold now, or what you select now.
Sovereignty lives in the future tense. It is the answer to a question you have not yet had to ask: when this provider fails me, what survives? If the honest answer is “everything essential, and I can take it elsewhere,” you are sovereign, whatever brands happen to sit on your desks today.
If the honest answer is “that depends on whether they let me,” then you are not sovereign, no matter how much you own or how freely you chose.
A closing note: this is not an abstract argument
Everything above can be stated in three letters and a contrast.
The OpenDocument Format (ODF) is what reimplementability looks like in practice. It is a published ISO standard that any developer can implement fully, without permission and without payment, and many independent applications do.
A file saved in ODF is not merely yours, it is re-creatable by anyone, which is precisely what makes it safe to entrust your institution’s memory to it. Nothing essential can be taken away, because nothing essential depends on a single vendor.
OOXML is the cautionary half of the contrast. It exists as a standard, and in that narrow sense a file saved in it will not “disappear.”
But the format as actually deployed carries provisions and behaviours that only its originator implements completely, and the effect is that full fidelity remains the property of one vendor. This is persistence without reimplementability, or the museum exhibit. You may keep the file forever and still be unable to do anything sovereign with it.
Which is why the most consequential decision a sovereignty-minded institution makes is not where its servers sit, nor which brand of software runs on them, but which format its documents are native in.
A platform that can export to an open format but lives, by default, in a proprietary one has granted its users persistence and quietly retained sovereignty for the vendor. The files will survive. The dependency survives with them.
Choose the native format that anyone can re-create, and the rest of the sovereignty question becomes manageable. Get that one choice wrong, and no amount of owned hardware or vendor-neutral procurement will buy back what has been signed away.
Nothing essential can be taken away. For documents, that sentence has a name, and the name is ODF.
This article was inspired by Roberto Di Cosmo’s formulation of digital sovereignty as the capacity to ensure that nothing essential can be taken away.