[en] The European Commission’s new regulatory framework, the Artificial Intelligence (AI) Act, has significant implications for the development of AI. The AI Act defines a set of strict requirements for high-risk AI systems, increasing the regulatory and compliance requirements on developers, providers, and importers of such systems. In this work, we present a comprehensive analysis of the effects of the key provisions of the AI Act on AI systems, and how federated learning, a machine learning paradigm gaining prominence due to its collaborative privacy-preserving approach, can mitigate these effects. We propose a Federated Regulatory Sandbox that eases the burden on developers by providing a way to train foundational models that facilitates compliance with regulations.
