Answers are grounded in this paper and cite their sources.
Objectives Artificial intelligence (AI) systems are increasingly used in healthcare for diagnostic, prognostic, and decision-support purposes. Their adaptive characteristics challenge established approaches to risk classification and conformity assessment. In the European Union (EU), medical AI is regulated under the Medical Device Regulation (MDR), the In Vitro Diagnostic Regulation (IVDR), and the Artificial Intelligence Act (AI Act). This scoping review examines how this dual regime shapes risk-adjusted conformity assessment for medical AI, particularly in relation to lifecycle dynamics that challenge static certification models. Methods The methodological approach followed the frameworks of Arksey and O’Malley and Levac et al. and was reported in accordance with PRISMA-ScR. Bibliographic databases and grey literature were searched for EU-relevant publications on risk-adjusted conformity assessment of AI systems in healthcare. After screening and full-text review, 126 publications between 1 January 2022 and 15 October 2025 were included and analysed using thematic synthesis. Results The literature highlights recurring challenges in risk-adjusted conformity assessment, including interface issues between the AI Act and MDR/IVDR, broad high-risk classification for medical AI, and governance challenges for adaptive systems across the product lifecycle. Further themes include gaps between regulatory principles and operational practice, limited harmonised evidence metrics, capacity constraints for notified bodies, and unresolved questions regarding post-market evidence generation and accountability. Conclusion The reviewed literature generally characterises the EU regulatory framework for medical AI as comprehensive, while identifying variability and open questions in its practical application. This scoping review summarises discussions on proportionality, lifecycle governance, and operational standards in risk-adjusted conformity assessment. Public Interest Summary Artificial intelligence (AI) is increasingly used in healthcare. While these technologies can improve diagnosis, prediction, and clinical decision-making, they also introduce risks, as many AI systems learn from data and may change their behaviour over time. Existing medical device regulations were not originally developed for such dynamic technologies. This article reviews research published between January 2022 and October 2025 on how AI systems in healthcare are assessed for safety and regulatory compliance in the EU. A total of 126 publications were analysed. The findings show that, although the EU regulatory framework provides a comprehensive legal basis for medical AI, its practical implementation remains debated. The literature highlights challenges related to regulatory coordination, technical standards, and post-market monitoring across the AI lifecycle. For patients and clinicians, this means that unclear monitoring duties and responsibility boundaries may shift accountability to healthcare professionals without effective means to verify AI performance.
No takes yet. Share an insight, caveat, or question.
Reisinger et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: