Artificial Intelligence Law And Digitalisation In Law

The Certification Of High-Risk AI Systems

This chapter offers an analysis of the certification regime for high-risk AI systems under the EU AI Act. Although certification represents an exception within the overall structure of the AI Act, it functions as an important legal instrument: it determines market access, defines compliance obligations, and shapes the intersection of private ordering and public oversight. The chapter first reconstructs the classification logic of high-risk systems as set out in Article 6 AI Act. Two complementary pathways—one product-based, the other use-case driven—structure the risk taxonomy. However, this approach reveals inconsistencies. The reliance on existing Union harmonisation legislation can result in regulatory blind spots, particularly where third-party conformity assessments are absent despite evident fundamental rights risks. Second, the chapter distinguishes between internal and external conformity assessments, outlines the provider’s duties from technical documentation to CE marking, and discusses the legal status of the EU declaration of conformity. Finally, the chapter reflects on the normative ambivalence of the AI Act’s conformity assessment logic. On the one hand, it represents a functional extension of European product safety law into the digital domain. On the other, it delegates significant regulatory responsibility to private actors. The chapter concludes that the AI Act's certification regime illustrates the broader challenge of AI regulation: how to align formal legal structures with the fluid realities of algorithmic decision-making in high-risk domains.

Dimitrios Linardatos
DOI: 10.53478/TUBA.978-625-6110-66-3.ch06