The new EU Artificial Intelligence Act (AI Act, 2024/1689) adds a layer of regulatory requirements for medical device firms already governed by the Medical Device Regulation (MDR 2017/745). In general, any AI “system” that is part of a CE-marked medical device will be treated as high-risk under the AI Act if the following conditions are fulfilled (I) it serves as a safety component of the medical device OR the AI represents the device itself, AND (II) already requires third-party conformity assessment under the MDR/IVDR.
From 2 August 2027, such AI/ML-enabled medical devices must meet the AI Act’s high-risk AI system requirements in addition to those specified in the MDR.
Overlapping and New Compliance Requirements
In practice, the AI Act largely mirrors many MDR obligations, but with some additional burdens. Both regimes require manufacturers to maintain detailed technical documentation, a quality management system, risk management processes, conformity assessments and post-market surveillance (referred to as ‘Post-Market Monitoring’).
The AI Act adds AI‑system specific requirements: for example, organizations providing AI Systems (AI system ‘providers’) must explicitly consider fundamental rights risks (privacy, non-discrimination, human dignity, etc.) and on top evaluate risks with regards to algorithmic bias and potential model drift. It further adds requirements for human oversight and enabling qualified operators to monitor or override and or halt the AI system), comprehensive data set validation (Article 10) requirements and automated log generation requirements.
In short, manufacturers will need to perform a gap analysis to identify and address any compliance obligations beyond those already satisfied under the MDR.
Ambiguities and Dual-Compliance Tensions
The AI Act’s concept of a “safety component” does not perfectly map to medical device terminology, creating further legal uncertainty over when AI itself is considered the medical device, or when AI is considered a safety component, or neither of these two options.
Conformity assessments also face a dual-burden issue. Going forward, a single notified body must perform a combined MDR/AIA assessment for high-risk AI medical devices. Yet reports warn that not all existing MDR Notified Bodies are pursuing AIA designation, and national processes to accredit AI conformity assessors are lagging. Manufacturers may be forced to switch to a new Notified Body that is AIA-notified – a potentially costly and time-consuming step. In short, firms worry that capacity bottlenecks and procedural misalignment will slow market access for AI-enabled devices.
Updated MDCG Guidance
In June 2025, the MDCG published guidance (MDCG 2025-6) clarifying how the AI Act interfaces with the MDR and IVDR. The document addresses several areas of regulatory overlap, including the status of clinical investigations and performance studies. It confirms that, for high-risk medical device AI, clinical investigations conducted in accordance with EU medical device laws (MDR/IVDR) are recognised as “real-world testing” under Article 60 of the AI Act, provided all MDR/IVDR requirements are met.
While the guidance resolves some key concerns, industry groups and notified bodies continue to call for more comprehensive rules or even targeted amendments to fully align the two frameworks.
Conclusion
Medtech industry and regulatory experts have voiced frustration at the dual-compliance burden. At a recent industry conference, stakeholders highlighted difficulties such as finding qualified AI specialists and interpreting overlapping obligations. Many have urged the EU to provide explicit guidance.
Medical device manufacturers and regulators alike are watching closely for forthcoming guidance and standards, hoping for more clarity on how AI Act obligations dovetail with MDR requirements.
