Announced at VivaTech, the partnership between Mistral AI and Nvidia is a step towards Europe’s ambitions for technological sovereignty. Finally, Europe is giving itself the means to compete in the AI race. Neither liberal like the US nor authoritarian like China, Europe wants to offer a ‘third way’. But will it be enough to enable the old continent to carry weight on the international stage?
To fully understand the European strategy, it must be seen as a three-pronged approach, where regulation and investment are designed to reinforce each other. On the one hand, regulation frames the market to build trust. On the other, investment programmes are being rolled out to stimulate the ecosystem and bring out the European champions that this regulation is supposed to govern.
The European strategy: regulation and investment
First propeller: the Artificial Intelligence Act (AI Act), which is the world’s first comprehensive horizontal legislation on AI. It imposes obligations proportionate to the potential risk of AI to citizens. Its main objective is to ensure that AI systems developed and deployed on the European market are safe, transparent and law-abiding in order to build trust: a prerequisite for their adoption by businesses and individuals.
Aware that regulation alone cannot build a competitive ecosystem, the EU launched an ambitious plan in April 2025: the AI Continent Action Plan. More than €1 billion per year in public funding will be mobilised through its two flagship programmes: Horizon Europe (€95.5 billion for the period 2021-2027) for research and innovation, and Digital Europe (€7.5 billion between 2021 and 2027) for deployment. The EU is investing on a large scale, not to mention the individual investment strategies of the various countries, such as the £109 billion planned by France.
However, Europe lacks tech giants and lags far behind in terms of private investment and computing power, with its cloud market dominated by US players (nearly 70%). The Commission itself recognises this innovation deficit and proposes to integrate support mechanisms, particularly for SMEs, such as regulatory sandboxes,AI Factories, which should eventually triple the continent’s computing capacity, and AI Gigafactories, dedicated to developing next-generation AI models that comply with European values and are financed by a private-public partnership.
The Brussels effect: geopolitical leverage
The Brussels effect is the third prong. This concept, theorised by law professor Anu Bradford, describes the EU’s ability to export its regulatory standards to the rest of the world, not by force, but by the attraction of its single market. Europe is building on this doctrine: making trust an export product, a competitive advantage and a global standard, just as the GDPR has become for privacy protection. With more than 450 million consumers with high purchasing power, the European market is essential for multinational companies. And to access it, they will be forced to adopt European standards, which are often the most demanding in the world.
This effect is itself based on two mechanisms: the de facto effect and the de jure effect. In the first case, multinational companies realise that it is simpler and more profitable to adopt the European standard for all their global operations, rather than managing different production lines, services or compliance policies for each region. European regulations become the de facto standard. Once these global companies have integrated costly European standards into their processes, they have a commercial interest in ensuring that their national competitors are subject to the same rules. They then start to put pressure on their own governments to adopt legislation similar to that of the EU in order to restore fair competition in their national markets.
A risky strategy?
Thanks to the Brussels effect, the EU is becoming a global rule-maker and can export its regulatory model and values. At least on paper. For this effect to occur, several conditions must be met: the size and importance of the EU market, its ability to develop and enforce complex and strict regulations, and above all, the fact that the object of regulation is inelastic and indivisible.
Except that the EU’s gamble is based on the assumption that the condition of ‘indivisibility’ applies to AI as it did to personal data. For the GDPR, it made more economic sense for a company like Facebook to unify its privacy policy globally than to manage fragmented data systems. But AI is a rapidly evolving technology. Regulating an adaptive algorithm is not the same as standardising the safety of a finished physical product. Anu Bradford herself has expressed doubts about the ability of the Brussels effect to apply so easily to such a dynamic and cross-cutting technology.
From a technical and economic standpoint, it is also conceivable for a developer to offer a high-performance AI model that does not comply with the requirements of the AI Act for the global market, and a restricted, less powerful but compliant version specifically for Europe. If the performance or cost differential between the ‘EU-compliant’ version and the ‘global’ version becomes too significant, companies will have a strong economic incentive to choose the global version. This would create a two-tier development ecosystem, with cutting-edge innovation taking place outside the EU’s borders. In this scenario, the Brussels effect would be neutralised.
Today, unlike the GDPR, which has inspired data protection laws around the world, the AI Act is struggling to win support. Allied democracies such as the United Kingdom and New Zealand have explicitly chosen not to follow the European path, opting instead for more agile approaches that are considered more conducive to innovation. And the few countries that are drawing inspiration from it, such as Canada and Brazil, are encountering difficulties in finalising their own legislation.
This is all the more true given that the AI Act is arriving in a highly competitive regulatory landscape, whereas the GDPR emerged in 2016 in a more or less unregulated environment. The United States is actively promoting its own, more flexible approach, embodied in the NIST’s AI Risk Management Framework. For its part, China is not hesitating to export its model of ‘digital authoritarianism’ through its economic partnerships. The European approach is therefore not the only option on the table, but one among many in an international regulatory battle.
Its effect Brussels could even backfire. This regulatory ‘caution’ could leave Europe ‘behind the innovation curve’, hamper the growth of its own tech champions and, paradoxically, increase its dependence on American and Chinese technologies.
So, is this the right strategy? It is still too early to give a definitive answer, but if it is not the best, it is at least the most consistent with the EU’s structural strengths and weaknesses. It may even be the only viable path for a power that cannot hope to compete with the United States and China on the sole grounds of raw innovation and venture capital. However, the success of this gamble hinges on two major challenges. The first is external: the ability of the Brussels Effect to establish itself in the field of AI, which is by no means guaranteed. The second is internal: the EU’s ability to reconcile its desire to regulate with its pressing need to innovate, without one stifling the other.
By Louis Chouvet
