A European geopolitics for artificial intelligence

In Europe, the situation is very different. The AI Act represents the most advanced attempt to build a regulatory framework that combines innovation and the protection of fundamental rights. It is an important achievement, based on the principles of transparency, proportionality and non-discrimination. But it cannot stand alone. Without a real common industrial policy, without an adequately funded public research system, and without a shared infrastructure, the Union runs the risk of remaining a regulatory power without real technological sovereignty.

The recent UN report emphasises that AI is not a neutral technology: it reflects the values and priorities of those who design it. That is why strong public governance is needed, capable of ensuring transparency, democratic control, and above all autonomy in the design and use of intelligent models.

Europe has some decisive strengths: a person-centred approach, a mature legal framework, and a tradition of research excellence. But it also has structural weaknesses: fragmented decision-making, delayed investment, lack of a common AI foundation capable of generating European language models at scale.

The OpenAI for Countries project thus forces us to look beyond the rhetoric of algorithmic accountability and ask an uncomfortable question: what space is left for democratic sovereignty when technical infrastructure is outsourced? The OpenAI for Countries project thus forces us to look beyond the rhetoric of algorithmic accountability and ask an uncomfortable question: what space is left for democratic sovereignty when technical infrastructure is outsourced?

The answer cannot be an ideological rejection of collaboration with the private sector. But neither can it consist in a silent surrender to the logic of extraction and centralisation. What is needed instead is a third, typically European, way that enhances the public-private alliance from a constitutionally oriented perspective.

In this perspective, a crucial role could be played by co-regulation, understood as a hybrid governance instrument that combines the technical dynamism of the private sector with the democratic legitimacy of the public. It is not a blank delegation, nor a form of self-regulation in disguise, but an operational model in which standards are constructed in a participatory manner, metrics are shared, and verification mechanisms are entrusted to independent third parties.

Two paradigmatic examples of this approach are the Code of Practice on Disinformation and the more recent Code of Conduct on General Purpose AI (GPAI), both promoted by the European Commission. In both cases, these are instruments that recognise the strategic value of private technological know-how, but hinge it within a publicist logic of responsibility, transparency and accountability. They are not mere 'good practices', but pieces of a broader strategy, which aims at building an algorithmic governance based on cooperation, but inspired by purposes of general interest.

In an ever-evolving technological environment, where the speed of innovation risks outstripping the speed of law, co-regulation may represent the European way to guard cognitive infrastructures, while guaranteeing effectiveness, traceability and compatibility with fundamental rights. Provided, however, that the primacy of public decision is maintained and that co-regulatory instruments are embedded in a legal ecosystem that is clear, verifiable and open to democratic evaluation.

Only in this way will Europe be able to play a leading role, and not a supporting role, in the new geopolitics of artificial intelligence. It is not enough to regulate: it is necessary to build. And to do so under the banner of democracy, competence and institutional farsightedness.