Talk by Silja Voeneky and Luca Cavalli

Rapid advances in artificial intelligence have made the effective regulation of high-risk AI models and AI-driven products and services increasingly urgent. Even where the probability of catastrophic harm remains uncertain or comparatively low, the potentially extreme magnitude of such harm calls for proportionate preventive action.

Silja Voeneky spells out in her presentation, why international human rights law provides a normative basis for duties of states to assess and mitigate AI-based catastrophic and existential risks. Against this background, the talk critically examines the regulation of high-risk AI according to the EU AI Act. Although the EU AI Act represents a first step towards comprehensive AI governance, significant gaps remain, including the absence of a permit procedure for high-risk AI systems.

The talk explores how high-risk AI models and AI-driven products and services can be governed through binding national, European and international law. A particular focus will be placed on adaptive regulation. Building on a regulatory model developed by Thorsten Schmidt and Silja Voeneky published in 2022, Voeneky proposes a risk-sensitive framework for high-risk AI-driven products and services. Companies and/or states could be required to provide proportionate “regulatory capital” before the market entry of a high-risk AI model, to be held in a fund as financial security against future AI-related harm. The broader aim is to spell out pathways to develop regulatory structures capable of keeping pace with rapidly evolving AI models while preserving innovation and ensuring that potentially catastrophic risks are effectively constrained.

More information to follow.