Every nation wants control over its digital future. The question is whether control means separation

A phrase has entered international policy debate with remarkable speed: digital sovereignty. Governments across every continent now speak of controlling their data, their infrastructure, their algorithms, their digital destiny. The impulse behind it is legitimate and, in many cases, overdue.

For decades, a small number of firms in a small number of countries have built the systems on which the rest of the world depends. Data generated in Lagos, Manila, or Nairobi is processed elsewhere, monetised elsewhere, and governed by rules written elsewhere. Nations have watched their citizens’ information become someone else’s asset, and their public services become dependent on infrastructure they neither own nor understand. That a country would want authority over this is not paranoia. It is a reasonable response to a genuine asymmetry.

But the word sovereignty carries an ambiguity that matters enormously for what happens next. It can mean the capacity to govern one’s own affairs. It can also mean separation from everyone else. These are not the same thing, and the difference will shape whether artificial intelligence is governed well or badly in the coming decade.

The Fragmentation Risk

Consider what happens if sovereignty is interpreted as isolation.

Each nation or bloc builds its own regulatory regime, its own data-localisation requirements, its own standards for what counts as safe. Compliance becomes so costly that only the largest firms can operate everywhere, which entrenches precisely the concentration that sovereignty was meant to challenge. Smaller nations, lacking the market size to make anyone comply, find their rules ignored rather than respected.

Meanwhile the harms themselves refuse to respect borders. A model trained in one jurisdiction is deployed in fifty. Disinformation crosses frontiers at the speed of a share button. A biased hiring system exported as software carries its bias into every labour market it touches. Climate costs of computation are paid by communities far from the data centres. A purely national approach to governing these things is like building a flood wall around one house on a floodplain.

There is a further danger, less discussed. Sovereignty framed as competition invites a race, and races have logics of their own. When nations understand AI primarily as strategic advantage over rivals, safety becomes a cost that slows you down, and caution becomes a weakness your competitor will exploit. The pressure runs consistently toward deploying faster and asking fewer questions.

What Cooperation Actually Requires

Yet the answer is not to dissolve sovereignty into a single global framework administered from a few capitals. That would reproduce the original problem in new clothing: rules made elsewhere, applied to everyone, in the name of universality.

The Catholic social tradition has a precise concept for navigating exactly this tension, and it is worth borrowing even by those who do not share its theological premises. Subsidiarity holds that decisions should be made at the level closest to the people affected, while higher levels of authority provide support when a problem exceeds what that level can handle alone.

Applied here, subsidiarity says two things at once, and both are necessary.

First, communities and nations have genuine authority over their own digital lives. A country is entitled to decide how its citizens’ health data is handled, what languages its public systems must serve, and which applications are incompatible with its values. This authority is not a courtesy extended by more powerful states. It belongs to them.

Second, no nation can govern alone what no nation contains. Standards for model safety, mechanisms for cross-border accountability, shared research on harms, coordinated responses to systems that fail in multiple countries at once: these require cooperation, because the alternative is not national control but national helplessness dressed as independence.

Subsidiarity resists both centralisation and fragmentation. That is exactly the balance the AI governance debate needs and rarely achieves.

Sovereignty Without Capacity Is a Word

There is a harder question underneath the diplomatic language, and it deserves naming plainly.

A nation cannot govern what it cannot build, understand, or audit. Digital sovereignty asserted by a country with no domestic compute, no local research capacity, no engineers trained to inspect a model, and no data centres within its borders is a legal claim without a technical foundation. The law says the data belongs here. The infrastructure says otherwise.

This is why cooperation and sovereignty are not opposites for most of the world. They are prerequisites for one another. Genuine sovereignty for the Global South requires capacity: universities that train researchers who stay, compute infrastructure that is locally accessible, datasets that represent local languages and histories, and the negotiating position that comes from having something to offer rather than only something to protect.

That capacity will not appear through isolation. It is built through partnership, technology transfer that actually transfers, research collaboration that treats African and Asian institutions as peers rather than data sources, and international arrangements that distribute capability rather than concentrating it further.

The test of any cooperative framework is therefore simple. Does it leave participating nations more capable of governing themselves, or more dependent on those who wrote the rules? Cooperation that deepens dependency is not cooperation. It is the old arrangement with better manners.

The Anthropological Question Beneath the Political One

All of this rests on a prior question that policy language tends to obscure: what is sovereignty ultimately for?

If sovereignty is about state power, then the debate is simply about which authority controls which systems, and citizens remain the object of governance rather than its subject. A government can achieve complete digital sovereignty and use it to surveil its own population more effectively than any foreign firm managed. National control is not automatically a gain for the people who live under it.

The deeper justification for sovereignty is that people should have a say in the systems that shape their lives, and that this say is most real when exercised close to home, in a language they speak, through institutions they can actually reach. Sovereignty matters because persons matter. It is a means, not an end.

This reframing has practical consequences. It means a sovereignty framework should be judged by whether ordinary citizens gain the ability to contest an automated decision, not only by whether servers sit within national borders. It means the participation of civil society, universities, faith communities, workers and affected populations in setting digital policy is not decoration but the substance of the thing. It means transparency to citizens matters as much as independence from foreign firms.

Solidarity adds the final measure. A framework that secures the digital future of wealthy nations while leaving poorer ones as suppliers of data and labour has not solved the governance problem. It has relocated it.

Cooperation as the Realistic Option

None of this is idealism about international relations. States pursue their interests, negotiations are difficult, and cooperative arrangements are slow and often disappointing.

But the alternative is worse on its own terms. Fragmentation produces a world where harms cross borders freely and remedies do not. Competition produces a race in which safety is the first casualty. Isolation produces the appearance of control without its substance, particularly for nations without the capacity to make isolation mean anything.

Cooperation is not the generous option. It is the realistic one. A country that helps build shared standards has influence over them. A country that withdraws has sovereignty over a smaller and smaller share of what actually affects its citizens.

The future of AI governance will be decided by whether nations can hold two things together: real authority over their own digital lives, and real willingness to build what none of them can build alone. That is not a compromise between sovereignty and cooperation. It is what sovereignty looks like when it is serious.