Paradigms of AI Regulation in the EU and beyond

The evolution of digital technologies comes with new regulatory challenges, which are even exacerbated by the emergence of Artificial Intelligence. This paper analyses the substance of the proposed EU AI Act, as well as contributions to the debate around it, to identify traces of different regulatory paradigms: a liberal, a public control, and a procedural paradigm. The liberal paradigm has been prevalent as much of the regulation of digital technology that was developed over the last decade or two was inspired either by the model of product regulation (certainly in the EU) or by the more specific and somewhat distinct model of intellectual property regulation. In both cases, this meant that regulation mostly aimed to protect consumers and to prevent excessive and harmful use of the technology (Benthall & Goldenfein 2020). Such regulation is unlikely to take account of the society-wide effects that digital technology may have as it structurally transforms the ways citizens interact, the information they consume, and the beliefs that they hold. To address such societal effects, short of putting technology under direct public control, regulators have imposed (self-)regulatory responsibilities on digital platforms, as is for instance illustrated by parts of the EU Digital Services Act. With the emergence of Artificial Intelligence these challenges are taken to a next level, since the inherently transformative character of this technology prevents substantive regulation. Instead, regulators rely on a mix of procedural obligations of technology developers and deployers in terms of disclosure, registration, licensing, and audits (Guha et al. 2023). To further tease out the distinctive EU approach to digital regulation, the paper contrasts its findings with the main characteristics of AI regulation in the US (specifically, President Biden’s Executive Order on Safe, Secure, and Trustworthy Artificial Intelligence) and in China (based on Sheehan 2023).