While transparency is often championed as the key to addressing the risks of automated decision-making (ADM) in public governance, this op-ed argues that a narrow focus on explainability overlooks deeper systemic issues such as power imbalances, commercial influence, and weakened accountability. To address these issues, mechanisms that promote transparency must operate alongside efforts to enhance citizen engagement and other methods of oversight and accountability to better protect democratic values.

Article 10 of the EU’s AI Act puts data governance at the heart of bias mitigation in high-risk AI systems but offers little guidance on implementation. Delegating these challenges to technical standardisation bodies raises both feasibility and legitimacy concerns, posing a significant test for the institutions now tasked with defining AI fairness in practice.