This paper examines the regulatory framework governing AI-based diagnostic systems under EU law. It analyzes how such systems are classified as high-risk AI under the EU Artificial Intelligence Act — a classification that derives directly from their risk classification under the Medical Device Regulation — and what requirements apply to their market approval. The paper further shows that the two frameworks do not operate in parallel: the AI Act is absorbed into the existing MDR conformity assessment procedure, adding AI-specific requirements addressing data governance, human oversight, and continuous learning without creating redundant compliance obligations. Primary sources include the EU AI Act, the MDR, and the joint guidance document MDCG 2025-6.
Table of Contents
1. Introduction
2. Terminology
3. Classification of AI Diagnostic Systems
3.1 According to the AIA
3.2 According to the MDR
4. The interaction between the AIA and MDR
5. Conclusion
Research Objectives and Key Topics
This study investigates the regulatory interplay between the EU Artificial Intelligence Act (AIA) and the Medical Device Regulation (MDR) regarding AI-powered diagnostic systems. It aims to clarify how these two frameworks coexist, how conformity assessments are integrated, and how the challenge of continuous machine learning model updates is managed within existing legal boundaries.
- Regulatory classification of AI systems under AIA and MDR.
- Integration of conformity assessment processes for medical AI.
- Management of dynamic software changes and continuous learning.
- Requirement alignment for data governance and system interpretability.
- Definition of roles and terminology for medical AI providers.
Excerpt from the Book
The interaction between the AIA and MDR
Every Medical Device (MD), regardless of the incorporation of AI, needs to undergo a conformity assessment by the manufacturer (only Class I) or a notified body [3]. The whole assessment process for Non-AI MDs was thoroughly lined out in [5]. One noteably finding of this study was, that even though all MDs need to comply with the the same General Safety and Performance Requirements (GSPR), the applicability of these requirements and how thoroughly compliance needs to be demonstrated depends not only on the risk class of the device, but also on the type of the MD. As a concrete example, the demonstration of electrical safety is only required for active devices (devices that use energy for their intended purpose).
According to Article 43 (3) of the AIA, the necessary conformity assessment for MDAI required by the AIA shall be integrated in the conformity assessment process already established by the MDR and guidelines of the Medical Device Coordination Group (MDCG) [2, 1]. In practice this means that manufacturers should establish one unified Quality Management System (QMS) and Technical Documentation (TechDoc) for requirements of the MDR and those of the AIA. But even though the conformity assessment process of the MDR integrates the conformity assessment of the AIA, for AI systems there is an extra layer of regulatory requirements that gets added into this process.
Summary of Chapters
1. Introduction: This chapter introduces the potential of AI in healthcare and outlines the regulatory challenge of ensuring safety and efficacy under the dual framework of the AIA and MDR.
2. Terminology: This section clarifies the disparate definitions of roles, such as providers versus manufacturers, which are critical for legal responsibility in medical AI.
3. Classification of AI Diagnostic Systems: This part examines the risk-based classification approaches of both regulations and how MDAI systems are categorized based on their intended medical purpose.
4. The interaction between the AIA and MDR: This chapter analyzes how conformity assessments are unified and how the AIA introduces specific requirements for data governance and change management into the MDR’s existing structure.
5. Conclusion: The final section summarizes how the integration of these two frameworks avoids redundancy while addressing the unique, dynamic challenges posed by machine learning in medical diagnostics.
Keywords
Artificial Intelligence, Medical Device Regulation, EU AI Act, conformity assessment, MDAI, risk classification, machine learning, software, regulatory framework, healthcare diagnostics, data governance, change management.
Frequently Asked Questions
What is the primary focus of this work?
This work focuses on the regulatory landscape governing AI-powered medical diagnostic systems in the EU, specifically the interaction between the Artificial Intelligence Act and the Medical Device Regulation.
What are the central themes of the analysis?
The central themes include regulatory classification, the integration of conformity assessment procedures, and the management of software changes in dynamic AI systems.
What is the main research objective?
The main objective is to analyze how the AIA and MDR interact to ensure patient safety without creating unnecessary administrative burdens or redundant regulatory processes.
Which scientific method does the author employ?
The author uses a descriptive and analytical legal study, comparing regulatory texts and frameworks to demonstrate their practical application and synthesis.
What does the main body address?
The main body addresses terminology differences, classification logic, the integration of assessments, and specific requirements for data governance and interpretability in AI models.
Which keywords best describe the research?
The work is characterized by terms such as AI Act, MDR, medical diagnostics, regulatory compliance, risk-based regulation, and continuous learning.
How are pre-determined changes handled under the new regulation?
According to the AIA, pre-determined changes specified in the Technical Documentation during the initial assessment do not require a new conformity assessment upon implementation.
Why is the "lex specialis" concept relevant here?
It is relevant as the AIA seeks to build upon existing sectoral frameworks like the MDR, delegating domain-specific risks to that legislation rather than creating a completely parallel and independent system.
- Arbeit zitieren
- Marlon Müller (Autor:in), 2026, AI-Powered Medical Diagnostics under Dual Regulation, München, GRIN Verlag, https://www.grin.com/document/1723862
