The European Union's approach to regulating artificial intelligence (AI) has adopted a governance model based on risk prevention, while recognizing the need to avoid closing the economies of EU countries to the spread of this powerful driver of innovation. Building on the adoption of Regulation (EU) 2024/1689 (the AI Act) and the subsequent 2025 Commission Guidelines on prohibited AI practices, this paper examines the mechanisms through which the EU attempts to reconcile market integration, innovation, and the protection of fundamental rights through a multi-level regulatory architecture. The analysis demonstrates that the EU's approach reflects a broader strategy of managing uncertainty through differentiated risk categories, delegated expertise, and multi-level execution. The underlying choice is not to consider AI as an additional risk in a specific context requiring regulation, but to view it as an environment requiring risk reduction. Comparison with the Chinese regulatory approach, which focuses on national interests and threats to the social order, and with the American one, which at the federal level emphasizes corporate self-regulation, highlights how the European model differs in its focus on the rights of individuals and workers. This approach in the US has been reflected in the legislation of some states in very different ways. In the US, therefore, at the federal level, the logic of global competition prevails, while at the state level, the need for social protection is felt. The European Commission's 2025 guidelines strive to make a regulatory model understandable, yet it remains difficult to interpret. The logic of risk classification to be applied in concrete cases must navigate different levels of responsibility and use risk simulation tools. At the end of 2025, the European Commission presented a proposal that recognizes the complexity of the implementation process and aims to simplify procedures.
AI Risk Prevention Multilevel Governance in Europe compared to the two-tier US Regulation: Citizens' Rights or Global Power Competition?
Nicola Giannelli
2026
Abstract
The European Union's approach to regulating artificial intelligence (AI) has adopted a governance model based on risk prevention, while recognizing the need to avoid closing the economies of EU countries to the spread of this powerful driver of innovation. Building on the adoption of Regulation (EU) 2024/1689 (the AI Act) and the subsequent 2025 Commission Guidelines on prohibited AI practices, this paper examines the mechanisms through which the EU attempts to reconcile market integration, innovation, and the protection of fundamental rights through a multi-level regulatory architecture. The analysis demonstrates that the EU's approach reflects a broader strategy of managing uncertainty through differentiated risk categories, delegated expertise, and multi-level execution. The underlying choice is not to consider AI as an additional risk in a specific context requiring regulation, but to view it as an environment requiring risk reduction. Comparison with the Chinese regulatory approach, which focuses on national interests and threats to the social order, and with the American one, which at the federal level emphasizes corporate self-regulation, highlights how the European model differs in its focus on the rights of individuals and workers. This approach in the US has been reflected in the legislation of some states in very different ways. In the US, therefore, at the federal level, the logic of global competition prevails, while at the state level, the need for social protection is felt. The European Commission's 2025 guidelines strive to make a regulatory model understandable, yet it remains difficult to interpret. The logic of risk classification to be applied in concrete cases must navigate different levels of responsibility and use risk simulation tools. At the end of 2025, the European Commission presented a proposal that recognizes the complexity of the implementation process and aims to simplify procedures.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
