Artificial intelligence and European sovereignty

INSIGHTS AND RECOMMENDATIONS

Over the last few years, artificial intelligence (AI) has become a hot topic, not only in the media and in the political sphere, but also in the business world, to the point that people are talking about the “digital miracle” or the “AI revolution”. And it is true that AI, and generative AI in particular, has turned life upside down, both inside and outside work. This was the subject of the report published by SKEMA Publika in December 2024, entitled The digital transition: impacts and dependences, which analyses the impacts and dependences related to the digital transition and the transition to AI. The political and strategic importance of this new technology has also been illustrated by recent events reported in the context of the rivalry between China and the USA. In January 2025, a Chinese AI tool named DeepSeek turned the tables in the race for AI. The low cost of developing it (less than $6 million according to China) worried its American competitor, ChatGPT¹ For comparison, the development of ChatGPT-4 is said to have cost $80 million, with OpenAI investing a total of $3 billion across all its models².

As a result, Washington heralded “Stargate”, a joint project by SoftBank, OpenAI and Oracle to invest $500 billion to build the infrastructure required to develop the AI industry and win the battle with China³. Across the Atlantic, the European Union (EU) committed to investing €1.5 billion for the period 2018-2020, and then €1 billion per year between 2021 and 2027, in AI-related research and innovation⁴. A further step was taken towards a European digital doctrine in January 2025 when a collective of over 200 European tech companies announced the EuroStack project. The collective observed that Europe sat at a historic tipping point, and that there was an urgent need for the continent to move on from its current status as a “digital colony” of the USA. Consequently, the project set out a series of binding measures and incentives to federate Europe’s digital power by buying, selling and financing European solutions. The idea is to put in place a doctrine where the technology, the governance and the financing work together to invest in Europe to build and use a suite of digital infrastructure. One of the measures proposed was a “Buy European Act”, to support the growth of European digital businesses⁵. Comments from a number of different people suggest that Europe needs to step up its game in the digital arena. In a speech on 18 February 2025, during European Parliamentary Week 2025, Mario Draghi pointed out the long-term effects for Europe of being left behind. He said that eight of the ten main language models had been developed in the USA, and the other two in China. In addition, he underlined the need for an unprecedented degree of coordination between national governments and parliaments, the European Commission and the European Parliament. In Mr Draghi’s opinion, annual investments of €750-€800 billion are needed to revitalise the European economy as a whole. He recommended achieving this objective through a combination of EU instruments and more flexible use of state aid⁶. The AI race is of strategic importance to the question of political sovereignty. Europe was late making the shift to the internet in the 1990s and 2000s, and cannot afford to fall behind in the adoption of AI. This is the justification for the relevance of this policy paper, based on the academic contribution of Ludovic Dibiaggio, Lionel Nesta and Simone Vannuccini in “European sovereignty in artificial intelligence: A competence-based perspective”⁷.It is a continuation of a first policy paper, published by SKEMA Publika in April 2022, entitled “Artificial Intelligence: A Political Subject. The main world powers and the place for France among them” based on the article “Artificial Intelligence: Key Technologies and Actors”, by Ludovic Dibiaggio, Mohammed Keita and Lionel Nesta.

We will examine the differences between approaches to developing AI tools across the world and attempt to define the European approach, pinpointing its strengths and weaknesses. On the basis of this analysis, we will offer recommendations to prompt the development of European sovereignty in the AI sector, against a backdrop of fierce competition between the USA and China.

In the face of the competition between these two powers, the EU has no option but to become a major player in AI in order to strengthen its economic competitiveness, through expertise in strategic technologies, and reduce its dependence on third countries. Currently, the European AI industry depends on American companies for the infrastructure, software and services it needs to develop AI models. Given the lack of reliable protection against intrusions and confidentiality breaches, this dependence is a real barrier to the development of a sovereign, independent AI. In this policy paper, we will set out the theory that, when considering the question of AI value chain integration and sovereignty it is important to take into account Europe’s dependence on foreign infrastructure, for example for semiconductors and cloud services. Indeed, without a commanding knowledge of this digital infrastructure, AI sovereignty can be no more than wishful thinking.

To evaluate the capacity of the EU to achieve technological sovereignty in the field of AI, the authors carried out an in-depth analysis of data on publications and patents by the major global players between 1990 and 2021. Working from this, they proposed a methodology for measuring innovation based on analysing capacity for integration and specialisation and on the levels and forms of investment. They have thereby succeeded in identifying the strengths and weaknesses of the European AI industry, which has become the subject of international rivalry and of an emerging “arms race” to steer and control its development.

This policy paper aims to give public- and private-sector decision-makers a comparative perspective on the strategic positions adopted by the leading AI countries. This information is essential to making informed decisions about investment, structure and collaboration.

SUMMARY OF RECOMMENDATIONS

Develop a common understanding of the direction of investment to promote the emergence of European champions in service industries, in the healthcare or culture sectors for example.
Collect and analyse data at EU level without breaching data privacy laws. To achieve this, it will be necessary to harmonise rules and data, following on from what has already been achieved with the GDPR.
Develop a strategy of integrating AI throughout the entire value chain, creating differentiation compared to the American and Chinese models by building an AI that is people-centred and trustworthy.
Guide the development of AI towards a specific productive specialisation. Define, with member states, the techniques that would contribute to the development of functions that could be used in the industrial or service applications in which Europe is dominant or wishes to gain market share. In doing so, take into account the importance of flexibility to adapt to new challenges and new technologies. One push strategy that could be useful would be a science- and university-focused approach, with funding directed towards the solutions needed by Europe’s industrial companies. The authors see public procurement as another potentially effective tool for stimulating innovation in general, and AI innovation in particular.
Attract the best staff thanks to higher salaries and better working conditions.
Put in place a clear regulatory framework that encourages innovation while promoting security and ethics by adopting the smart law concept. This concept would demonstrate a balance between innovation, security and ethics by providing a pragmatic framework for AI.