Between innovation and regulation, what choices for Europe?
Conferences-debates organized by the Debouzy firm, February 12, 2025.
In the aftermath of the “AI Action” summit organized in Paris and the “open house” day at Station F, as well as the publication of the Draghi and Letta reports and the European Commission’s “Competitiveness Compass” plan, but above all, the election of Donald Trump and the ongoing geopolitical and technological upheavals, what will be Europe’s place in the new balance of power? Can it impose its rules as a global reference? Does the regulation advocated during the European Commission’s previous mandate – particularly in terms of personal data protection (GDPR) – have any chance of being implemented in AI?
Marc Mossé (environmental advisor at Debouzy) recalled the challenges of the Paris summit, which brought together heads of state and leaders of major tech companies. He noted that for the past 15 years, AI has no longer been a matter of regulating a “common good”, but has become a “lever of power” for a State in the confrontation between geopolitical blocs. He noted that since the emergence of generative AI in 2022, the financial stakes have increased tenfold, with the US raising more than $500 billion, the European Union more than €200 billion and France €109 billion (apparently not included in the European envelope), to finance the development of AI.
Pierre Sellal (French Ambassador) stated that the Paris Summit had made it possible to better measure the impacts of AI in certain sectors such as defense, health or energy, but that paradoxically, the effects of AI on employment and education, as well as the ethical questions posed by certain applications of AI, had not been sufficiently debated. He noted that major differences had emerged between the American and European approaches to AI regulation. Two major countries – the United States and the United Kingdom – refused to sign the final declaration of the summit.
Florence G. Sell (professor at Stanford) observed a widening of the cultural divide between American and European (especially French) jurists, the former being in favor of a soft law based on AI “governance” and a flexible framework of the AI value chain, and the latter being more attached to a hard law based on ethical principles and regulation of practices. The former are opposed to the “bureaucratization of AI development” (the “Brussels effect”) and the latter are hostile to libertarian practices of crony capitalism in the name of freedom of creation. GAFAM, however, is rather opposed to open source applications, preferring an open source way, which reveals the logic of data processing, but not the source codes themselves and the training databases considered as business secrets. The recent open-source release of the Chinese DeepSeek solution has sparked reactions in the AI ecosystem, with several experts believing that the software was trained on data from the results of GPT chat processing (“distillation”) and that this collective practice could in the future reduce the costs of storing and processing current databases.
Mahasti Razavi (partner at Debouzy) assures that the European Union – and France in particular – has many assets in the face of American and Chinese competitors. It has an elite group of AI engineers and several large data storage centers. As of 2024, more than 800 start-ups have been created in France. However, the European market for users of AI solutions remains too small and fragmented compared to the US and Chinese markets. Mahasti Razavi advocates the long-term establishment of a “security framework” based on trust between the multiple stakeholders of AI (combining experts, practitioners, politicians, lawyers and academics) and based on practicable rules of “good governance” applied to a “more frugal AI” than the current general AI. She recognizes that the current AI Act needs to be simplified.
The debates that followed the conferences focused in particular on the positive impacts (AI solutions can generate competitive advantages for companies) and negative impacts (AI destroys certain jobs and involves unprecedented retraining efforts, mainly in the service trades). In particular, efforts will have to be made in terms of teaching and learning new professions.
Notes by J-J. Pluchart
