Le piège de la souveraineté numérique

Ce texte en anglais restitue mon intervention lors du panel « What Digital Sovereignty Means for Civil Society » organisé par le chapitre suisse de l’Internet Society (ISOC) à Zurich le 27 mars 2026.

Today’s debate on digital sovereignty feels like an intense case of déjà vu: For the better part of the past thirty years, Europe’s key policy-makers have kept playing a familiar song to convey a sense of urgency and impose their latest stripe of industrial policy. That song goes something like this:

“Europe risks falling behind – behind the US, behind Japan, behind China. We need to catch up, we need a European Google [or replace by any other technology or overlord dominating contemporary capitalism].”

More often than not, this has meant deregulating, attracting foreign investment, relaxing fundamental rights protections.

Hyperscale to nowhere

This is exactly what is happening today in the arms race towards AI and massive investment in computing power. To achieve “digital sovereignty”, the European Commission has tabled the environmental and tech omnibus proposals, both of which follow suit on the 2024 Draghi report to boost AI infrastructure in Europe. And to do so, both enact historical rollbacks of fundamental rights and environmental protections, all in the name of extracting data to train AI models and fast-tracking the construction of data centers.

For the past year and a half in France, civil society groups have been fighting a similar reform, trying to counter Emmanuel Macron’s 109-billion-euro plan to build data centers all across France. In Paris and elsewhere in the country, the motto of digital sovereignty is almost always invoked to defend a proposal that strips down environmental regulation and public participation on major industrial projects.

But what kind of digital sovereignty are we really talking about?

Let’s start with the Paris-based Mistral AI, a startup that is usually presented as the guarantee of French and European technological sovereignty. What is less often mentioned is the fact that it counts several prominent US investors in its capital, including active supporters of Donald Trump such as Marc Andreessen and Ben Horowitz.

Then, among the 109 billion euros in announced investments in 35 “hyperscale” AI data centers across France, 50 billion come from the United Arab Emirates, and 20 billion from the Canadian fund Brookfield. Most of these mega-facilities will run the latest NVIDIA chips and US AI models – directly replicating and amplifying Europe’s structural dependence on Big Tech.

Represented in part by the France Data Center lobby group, companies like Amazon, Google and Microsoft have been quietly pushing this deregulation. Their interest in France seem mainly driven by their sense that France’s agressive nuclear policy can help contain their booming carbon emissions. In central and southern Europe, Big Tech also get closer to Africa, where they can compete with Chinese companies for digital hegemony on that other crucial continent.

Toxic fumes

Data centers are revealing objects, politically and materially. For decades, mainstream discourse treated the digital as weightless – a cloud, a flow, something immaterial. We now know – as we should have known then – that is a myth. Cyberspace is entirely material. It is land, water, energy, made into cables, chips, roaring servers.

I live in Marseille, which has become the seventh-largest hub at the global level in terms of the concentration of submarine cables and data centers. Over the past decade, several data centers have been built in the industrial harbor, whose spatial footprint grew alongside France’s colonial conquests.

Data centers in Marseille raise many issues. First, they deprive underserved neighborhoods of access to the seafront. More generally, like in many other places, these huge concrete warehouses prevent many other possible appropriation of that land. They also require enormous quantities of fresh water that could support urban agriculture in Marseille’s dry soils, but is instead used to cool off the servers before being expelled into the sea. When they don’t spoil clean water, server-cooling systems of the data center giant Digital Realty, the main data center builder and operator in Marseille, use fluorinated gases. Now, their greenhouse effect is hundreds of times more potent than CO₂ but that has not prevent negligence on the part of the company leading to several leaks. And then there is the most pressing issues, namely the fact that the rapid expansion of these gigantic warehouses have already led to the postponement of dock electrification, meaning cruise ships continue pumping toxic fumes over the Saint-Antoine district, where the prevalence of various diseases linked to air pollution is particularly high.

I could go on. Thousands of local communities across the world are dealing with a version of this same story. As the conflicts around data centers escalate, almost everywhere political and business elites resort to the sovereignty narrative to impose them. A reminiscence of the historically-entrenched connection between sovereignty, the modern state, territorial conquest, and colonial violence.

Acceleration

Europe’s AI policy and the dominating narrative on digital sovereignty are not only reckless – they are strategically absurd given that the AI race is likely already lost. We may put AI everywhere in the vain hope of not being left aside, in the name of innovation, progress, openness, sovereignty. But in the process, we will either deepen of our dependence on foreign Big Tech, or construct a pale imitations of it. A European Google or a European ChatGPT is still an extractivist, capitalist, surveillance machine.

Either way, by accelerating AI, Europe will keep digging the rabbit hole in which it already finds itself as AI will work alongside austerity measures and foster the taylorization of workers — including those who, so far, were largely shielded from it, that is the creative and intellectual classes. By winning over managers obsessed with productivity as well as users trapped by the “dictatorship of convenience”, it will further undermine quality of life and aggravate inequalities. It will consume immense financial resources that could fund genuinely different approaches – not to mention the natural resources extracted and burnt up in the process. By doing so, it will accelerate the drift towards new forms fascism.

Short of a major political shift, this is what the actualization of digital sovereignty policies will look like. Which may explain why, for all its reactionary outlook, the far right is perfectly happy to embrace the AI disruption. As Dan McQuillan, author of Resisting AI, notes, the idea of “sovereign digital infrastructure”, even where they might actually entrench dependence, resonates strongly with their ultranationalism. In France for instance, the National Rally far has been pushing strongly in favor of fast-tracking data center construction, which fits nicely with its program in favor of nuclear and other fossil-based energy sources.

Seen in that light, digital sovereignty is not a counter-project. In the current context, it risks being an accelerant towards a Europe’s particular blend of techno-fascism.

Digital Autonomy

Here I want to propose a conceptual distinction. The language of sovereignty – from Bodin through Hobbes to Rousseau – has always been about the concentration of ultimate, indivisible power: who commands, whose will prevails. What we need instead is the language of autonomy, in the sense the philosopher Cornelius Castoriadis gave it: a community capable of collectively reflecting on and transforming the norms and power structures that govern it. Not a sovereign center deciding for all, but an ongoing, deliberate, distributed self-institution.

The alternative to digital sovereignty is a different politics altogether. In the environmental justice movement, there has long been an idea of degrowth – de-escalating our ecological footprint. Crucially, as McQuillan and others argue, degrowth doesn’t stop with refusal. It addresses power structures. It is conducive to a techno-politics that challenges both AI acceleration and the growth imperative. It switches the focus to other ways of organizing, other ways of doing. Refusing the “sovereign AI” paradigm, then, can form the starting point of a discussion opening up on a radically different future towards greater autonomy.

This isn’t purely theoretical. Ten years ago, together with Panayotis Antoniadis who invited me to this event, as well as many other colleagues, we started a project called netCommons – an attempt to promote alternative ways of building digital infrastructure. One where communications networks and the services running on top of them are governed as commons, tailored to human needs, with governance democratized and costs and impacts handled as close as possible to the communities that bear them: community wireless networks, federated protocols, cooperative infrastructures.

Many communities across the world cultivate these alternative ways for making and relating digital technologies, from redesigning communications protocols to adapt to a resource-scarce world to lightweight content publishing tools. Crucially, many also fight to resist the imposition of digital technologies, cultivating ways of meeting one another and building deep connections without the mediation or diversion of digital screens. Some have even scored important victories towards the right to access essential services by connecting directly as humans rather than through the mediation of this complex digital apparatus (see, e.g., the “L’Humain d’abord” campaign in Belgium). Less machinery, less automation: this is what McQuillan calls “decomputing.”

These endeavors may seem to marginal. But they point to an alternative future. They remind us that the choice is not between Big Tech and a European replica of Big Tech.

Provocation

I want to close with a provocation. It is actually found in the pseudonymous preface to the French translation of philosopher Yuk Hui’s work, written under the name Junius Frey.

After having analyzed contemporary Chinese governmentality, the author introduces Alexandre Kojève’s 1945 “Sketch of a French Policy Doctrine”. Kojève – a French philosopher and civil servant, and probably one of the strangest characters in European modern history – starts from a Schmittian observation about the transformation of political units: nation-states, he argues, are ceasing to be genuine political realities. Instead, the age of Empires has arrived, and France and Europe were squeezed up by two rising colossi: the Anglo-American Empire, founded on the cult of work and of the economy, deploying an individualistic productivism; and the Slavo-Soviet Empire, of Orthodox inspiration, whose productivism was collectivist in nature.

Let’s leave the discussion on the descriptive accuracy aside. What is interesting is Kojève’s proposal: Drawing on a project already sketched by the Occitan resistance fighter Jean Cassou, Kojève proposed a “Latin Empire” – not necessarily as another industrial or military bloc, but as something closer to a civilizational counter-offer. Its foundation would lay in a shared mentality, one bound by…

… “that art of leisure that is the source of art in general, by the talent for creating that ‘sweetness of living’ that has nothing to do with material comfort, by that very ‘dolce farniente’ that only degenerates into laziness if it doesn’t follow upon a productive and fecund labor.”

Kojève then goes on to stress the “municipal” character of the “Latin World” and talks of the Renaissance as “the historical Latin period par excellence.”

It is an odd, half-forgotten text, and Frey is careful not to romanticize it. Still, he is struck by its relevance for our present moment. Today, industrial and military rivalry centers on digital and computational power. For Europe (and other regions), the trap is the same as that evidenced by Kojève: that of competing on terms set by empires you cannot and should not match. To escape that trap, you need to play a different game. Frey writes:

“If the current economic and technological course is headed straight for the wall, then it must be admitted that taking a step backward might mean taking several hits in advance. Deserting the always-already doomed game of the powers-that-be might inaugurate a new kind of game. Tearing oneself away from the historical struggle could be the only way to prevail over a confrontation that is itself a losing proposition. Let us leave ‘ChinaAmerica’ to its sad fate.”

The alternative to digital sovereignty as AI acceleration – with its adjacent techno-fascism – means refusing the arms race altogether, opening up our imaginaries to cultivating an “art of leisure”, a different relationship to time and space, a “sweetness of living” grounded in democratic, municipal forms of self-government. It may sound naive. It surely opens up tons of difficult questions. Yet, I would argue it is the most promising path to simultaneously address the ecological emergency, preserve democratic forms of life, and incidentally build truly emancipatory technical tools.