Harmonised Standards and Conformity Assessments in the AI Act: Strengthening Independent and Participatory Oversight

Abstract

Adopted by the European Union in 2024, the Artificial Intelligence Act (AIA) constitutes a landmark framework for the regulation of AI systems within the EU’s internal market. While recognising its significance for fostering accountability and trustworthy AI, this position paper focuses on two central mechanisms for its effective implementation: (1) technical standardisation and (2) conformity assessments. It argues that, in their current formulation, both mechanisms risk an excessive reliance on private governance, thereby constraining democratic oversight and the effective protection of fundamental rights. The AIA implements its legal obligations through technical standardisation under Article 40, delegating detailed requirements to the European standardisation organisations. Compliance with these harmonised standards creates a presumption of conformity, giving companies strong incentives to follow them. Although the AIA introduces broader stakeholder consultation via the AI Board and Advisory Forum, participation remains non-binding, allowing large corporations to dominate standard-setting and raising concerns about democratic legitimacy and fundamental rights protection. Under Article 43, conformity assessments verify whether high-risk AI systems meet AIA re- quirements before market entry. Most rely on internal self-assessment; external control is only required in very limited cases. While this reduces administrative burdens, it also risks weakening oversight, transparency, and protection of fundamental rights.