Anticipation practices of public sector organizations in liminal AI design spaces – the case of the EU AI Act

Governmental regulation shapes the development and use of technology. Artificial intelligence (AI) has attracted significant regulatory initiatives, such as the EU AI Act (AIA), which illustrate the substantial lag often between a law proposal’s initial publication and its entry into force. This creates a liminal space in which organizations must anticipate forthcoming rules: How will compliance requirements affect our future AI plans? For researchers studying regulation within the information systems (IS) field, this period is largely uncharted because regulations are typically treated as existing or absent, rather than processes. We propose that the liminal AI design space can serve as a theoretical lens to illuminate the in-between nature of anticipatory AI work in organizations while regulation is taking shape. This lens challenges the static treatment of the regulation of technology by focusing on the early phases of the policy process. Empirically, we examined nine Finnish public sector organizations that prepared for the AIA while it was still in the making. We contribute four sets of anticipation practices that organizations enact to navigate emerging liminality. These practices reflect the dynamic conditions under which they must prepare for future AI development and use amid regulation in the making.