A Post-Normal Science Framework for Rethinking EU AI Governance

Taking its vantage point from the Post-Normal Science (PNS) framework, this paper examines how the governance of artificial intelligence (AI) has become a test case for democratic societies confronting conditions where ‘facts are uncertain, values in dispute, stakes high, and decisions urgent.’ It traces how PNS, originally developed for other science-policy domains marked by inadequate expert-driven policymaking and contested categories of risk and safety, offers analytical and normative resources for reimagining AI governance. The framework’s core principles - embracing irreducible uncertainty and constituting extended peer communities - find contemporary expression in United Nations Educational, Scientific and Cultural Organization’s (UNESCO) participatory assessment methodologies which treat AI systems as sociotechnical assemblages requiring ethical scrutiny rather than merely technical certification. By contrast, the European Union (EU) AI Act, despite its ambitious scope, embeds a troubling contradiction: it recognizes AI’s dynamic and unpredictable character while operationalizing oversight through risk categories, conformity assessments, and industry-led standardization processes that assume knowability and control. Now that the EU AI Act has entered into force, the decisive arena for responsible and democratic AI governance has shifted from legislative debate to the seemingly quiet, procedural machinery of implementation through standards. Standardization emerges not as a neutral technical exercise but as a political process determining whose expertise matters, which harms register as regulable and what remains invisible to oversight. Here, corporate influence threatens to calcify into epistemic capture, encoding industry priorities as objective technical requirements. The paper argues that critical scholarship must engage these seemingly procedural spaces as sites where the material and epistemic foundations of rights and freedoms are actively being constructed - and where knowledge itself becomes both a source and outcome of law-making. Only through co-regulatory processes that embrace the ‘uncomfortable knowledge’ and epistemic humility demanded by PNS can Europe realize its stated ambition to steer AI’s trajectory through a distinctive and responsible model of governance.