AI in France: betting on AI adoption and sovereignty rather than racing for the most powerful models

By Lina Jahidi

Interview with Erwan Paitel, state administrator and international AI expert

France is now establishing itself as an undisputed leader in artificial intelligence. Ranked 2nd worldwide and 1st in Europe in the latest edition of the Government AI Readiness Index, it consolidated this position by hosting the Global AI Summit in Paris in 2025. Between the rise of Mistral AI, the evolution of the National Digital Council (CNNum) into the CIAN (National Council for AI and Digital), and an ecosystem of over 1,000 start-ups including 16 unicorns, France has become the leading destination for foreign AI investment in Europe.

To understand the inner workings of this success, we spoke with Erwan Paitel, state administrator, inspector general and international expert on the societal impact of AI. What strikes you immediately in his analysis is a firm rejection of short-termism: according to him, the French strategy is not a last-minute reaction. It is built on solid foundations, the AI for humanity strategy was one of the first published in the world, along with the ministerial commission report AI: our ambition for France (2023-2024), of which Erwan Paitel was a member. From digital sovereignty to citizen education, through energy and cultural challenges, here is what this strategy reveals in depth.

Sovereignty through the “service layer”

One of the great strengths of the French strategy lies in its clarity about which battles are worth fighting. Rather than exhausting itself solely in the race  for large language models, “an interesting one, but one that will come to an end” as Erwan emphasised, France is resolutely betting on service-based AI; meaning using AI models as tools to meet specific professionals’ needs for several sectors. It is at this application layer that true value lies today, and it is a terrain where competition is far more balanced than in the race for the colossal infrastructure of American giants. The French AI strategy can also be characterised by how its pursuit of sovereignty translates into concrete partnerships. The State invests in Mistral, the French Large Language Model developer, to support its growth; in return, the French flagship makes its model available to public services and the military. This financial and technological flow protects French companies from international takeovers and encourages private players to relocate their solutions. 

But sovereignty does not stop at models: it also runs through infrastructure. Before 2024, France had few data centres and was heavily dependent on American giants like Azure or AWS. This situation sparked a heated controversy around the Health Data Hub, as all French health data was stored on Microsoft servers to develop AI models.

To address this challenge, France introduced the SecNumCloud label, emphasises Erwan. This framework imposes strict requirements: data must be physically hosted on national territory. The strategic aspect of this approach is that it does not close the door to foreign players, Amazon or Microsoft can offer SecNumCloud services if they comply with French security rules. Sovereignty is no longer defined by the nationality of the company, but by its ability to meet State requirements. Pragmatic and open, this logic helps compensate for the lack of local funding while increasing computing capacity on French soil.

Beyond the technical and economic dimensions, sovereignty is also cultural. Global models tend to flatten information and weaken historical and linguistic specificities. France therefore champions the development of French-language models capable of preserving the richness of  shared heritage, so that algorithms do not only offer a standardised culture, but also highlight local identities, states Erwan Paitel.

If the question of sovereignty defines France’s strategic posture, it cannot be realised without another fundamental pillar: talent development.

Developing talent: beyond simple coding

On the question of training, Erwan Paitel immediately challenges conventional wisdom. The automatic response is usually to train large numbers of developers to generate code. Yet this vision is risky: generative AI has itself become highly capable of coding. We must therefore move beyond the instinctive reflex of believing that “being good at AI” means only “knowing how to code“. For Erwan, the challenge has shifted from building the tool to its practical application: “And that’s why, rather than being in a race to develop tools and LLMs […], it’s better to focus on the service layer”. 

The vision put forward by Erwan Paitel focuses on the service layer of artificial intelligence: understanding how to use AI for the good of all, and how to make it genuinely useful in everyday life. This requires training varied profiles, capable of mastering the tool but also of understanding regulation, law and governance. The goal is to create experts able to bridge AI and other disciplines, adapting this technology to every sector of society. 

Education in critical thinking is another major pillar. Artificial intelligence is now everywhere, particularly on social media, and it produces an enormous volume of information, sometimes including errors or hallucinations. The challenge is not only knowing how to use the tool, but understanding how it works in order to maintain an analytical perspective. This need is particularly urgent among younger generations: today, 85% of 18-24 year-olds use generative AI.

For Erwan Paitel, this education must begin in primary school, between the ages of 7 and 10, “before children even own their first smartphone”. The goal is to prepare them for a digital world where AI can become a powerful vehicle for manipulating opinion. But this awareness-raising does not concern only students: it extends to the entire population through initiatives such as “AI Cafés“. These community gatherings, organised in towns and villages across France, allow people to discuss technology in a concrete and accessible way, regardless of their level of education. Technological sovereignty, citizen education, transformation of professions: these three axes outline the contours of a coherent French ambition. 

But for it to be truly sustainable, it will still need to meet two challenges of a very different nature. The future of AI in France will depend largely on its ability to reconcile technological performance with human responsibility. Erwan Paitel identifies two major challenges that transcend economic competition: social connection and ecology. On the human side, AI must not replace relationships between people. Technology must remain in the service of human connection, not substitute it. On the environmental side, while France benefits from a real advantage thanks to its low-carbon electricity, AI remains a heavy consumer of water for server cooling and of rare metals. Ignoring these constraints would mean building a power on fragile foundations.

Therefore, according to Erwan Paitel, this is why France’s success will not be measured solely by its economic growth, but instead by its ability to integrate this technology without harming human life or the environment. In this sense, the French strategy embodies a deep conviction: AI is not a simple technological wave that will pass but it is a lasting elevation of the global technological level and it is up to each nation to choose how it wishes to respond. France has made its choice clearly: with ambition, pragmatism, and a constant concern for the human.