Fundamental Rights Impact Assessment as Human Rights Protection Tool under EU AI Act

Abstract

This thesis asks whether the Fundamental Rights Impact Assessment (FRIA) under Article 27 of Regulation (EU) 2024/1689 actually protects fundamental rights before high-risk AI is deployed, or whether it ends up as a paper exercise - and what design lessons follow for Ukraine. The study sets out the constitutional foundation of the AI Act under Article 114 TFEU and the New Legislative Framework, traces the relationship between high-risk AI deployment and the Charter of Fundamental Rights, and draws normative criteria from the Data Protection Impact Assessment under Article 35 GDPR and from the Human Rights Impact Assessment tradition. It then analyses Article 27 doctrinally and compares it against the HRIA and DPIA, the Dutch FRAIA, the Canadian Directive on Automated Decision-Making, and the Corporate Sustainability Due Diligence Directive. The central finding is that Article 27 is built in a way that pushes deployers toward box-ticking: it satisfies none of the four HRIA process criteria as binding legal requirements and only partly meets the content criteria. The thesis turns that diagnosis into a connected set of seven design choices for Ukraine, grounded in Articles 3 and 22 of the Constitution and tailored to Draft Law No. 8153, the National AI Concept, and wartime institutional capacity. It shows that approximation under Article 474 of the EU- Ukraine Association Agreement requires Ukraine to reach the same protective result as EU law, which permits (and in fact calls for) a Ukrainian FRIA stronger than its EU model.