Does governing artificial intelligence mean governing data? When looking at data (governance), what do we not see?

Abstract

This commentary responds to Taylor et al.'s proposal for global data and artificial intelligence (AI) justice principles. While the proposed data governance framework is timely and important, I offer two provocations to invite critical reflection on the limits of the data-centric approaches to governing AI and digital societies more broadly. Drawing on the findings from the ERC INFO-LEG project, first, I argue that framing digital problems primarily in terms of data risks oversimplification of such problems, which are often complex socio-technical and political affairs. Second, I caution that data governance and the ensuing uncertainties of the boundaries of included data and groups can distract from the essence of the problems at hand, facilitating evasion strategies for powerful actors, for example, by navigating technical and legal complexity to their advantage. I advocate for more context-specific approaches to governing digital societies, where data governance may be part of the solution but not the starting point.