Forget the Facts, Aim for the Rights! On the Obsolescence of Empirical Knowledge in Defining the Risk/Rights-Based Approach to AI Regulation in the European Union

This article critically examines the inception of the recent European Commission (EC) proposal for a regulation laying down harmonization rules for Artificial Intelligence (Al Act). By establishing a four-level taxonomy of AI-related risks (non-high, limited, high, unacceptable) and corresponding technical standards, this instrument aims at preventing the occurrence of risks caused, in particular, by so-called high-risk Al systems. Though by virtue of its purpose and design, the Al Act follows a risk-based approach to regulation, it displays a specificity when compared to existing risk regulation in the European Union (EU), in such areas as environment and health. This specificity stems from the operative definition of risk the Al Act relies on: the risks covered in this proposal are not scientifically measurable threats of physical harm but threats of human rights violations which are difficult to quantify. In light of this, this article raises the issue of the evidence, if any, the EC gathered in view of designing a proportionate (i.e., reality conform) regulatory framework on Al. Through a critical analysis of the discovery procedures (public consultations) the EC launched for the purpose of drafting the AI Act, this article finds that the Commission disregarded the evidence gathered on several important points, namely the type (horizontal or sectoral) of regulation best suited for AI-related risk prevention. Considering the limited impact of empirical knowledge on the design of the Al Act, this article seeks to determine if, in lieu of being a normative response to factual reality, the instrument under consideration gives specific normative expression to fundamental values and rights, as well as to strategic objectives (trust and excellence) associated with the EU's Digital Single Market. This analysis allows to uncover two key standards having served as referents for the drafting of the Al Act: its axiological congruence with overarching Union values and its consistency with corresponding and already existing EU regulatory instruments. By delving into the origins of the Al Act, this article provides a unique insight into its normative rationale, uncovering the reasons for the relative obsolescence of empirical knowledge in its design and shedding more light on its specific legal nature as both a rights-protecting and risk-regulating instrument.