The Governance of Generative AI: Three Conditions for Research and Policy

Abstract

The increasing permeation of society by generative AI systems like ChatGPT has given rise to a pressing task that remains unresolved: the design of future-proof governance mechanisms that ensure democratic oversight over those AI systems. To establish and examine this oversight, it is essential that generative AI systems can be opened up for regulatory scrutiny. This chapter argues that there are three overarCHing dimensions to structure research and policy agendas about the governance of generative AI systems: analytical observability, public inspectability, and technical modifiability. Empirically, the chapter explicates those conditions with a focus on the EU’s Artificial Intelligence Act (AI Act). Those three conditions act as benchmarks to help perceive generative AI systems as negotiable objects, rather than viewing them as inevitable forces.

Ähnliche Arbeiten