The European Union's latest attempt to streamline regulation for AI-enabled medical devices may be creating as many problems as it solves. A December 2025 proposal to reform the Medical Device Regulation—dubbed MDR 2.0—aims to reduce regulatory burdens and eliminate double regulation for AI medical technologies, but a multidisciplinary analysis published in npj Digital Medicine suggests the reforms leave critical challenges unresolved and may actually increase uncertainty at crucial classification boundaries.

The stakes are high: manufacturers are already reporting product launch cancellations, device discontinuations, and considerations to exit the EU market entirely under current regulatory pressures. Yet the proposed solution presents a complex landscape where companies must navigate multiple competing regulatory pathways with unclear timelines extending through 2028.

The Double Regulation Dilemma

At the heart of the reform lies a fundamental problem: AI-enabled medical devices classified as high-risk under the AI Act—those at Class IIa or higher requiring Notified Body involvement—currently must meet requirements from both the Medical Device Regulation and AI Act frameworks simultaneously. This creates redundant compliance burdens that the European Commission sought to address.

The MDR 2.0 proposal attempts to resolve this by moving the MDR and In Vitro Diagnostic Medical Devices Regulation from Section A to Section B of AI Act Annex I, effectively removing direct application of AI Act Chapter III obligations to medical devices. However, this approach competes with an alternative pathway: the Digital Omnibus on AI (Regulation EU 2026/1744), adopted in July 2026, which delays high-risk requirements until August 2028 and requires the Commission to adopt delegated acts by August 2, 2027.

The result is regulatory uncertainty about which pathway will ultimately prevail, leaving manufacturers in limbo about compliance strategies.

Classification Confusion at the Boundaries

Perhaps more troubling, the revised classification system may introduce new ambiguities rather than resolving existing ones. The proposal revises Rule 11 to make medical device software Class I by default, with up-classification when output informs or drives clinical management in serious situations.

Yet the analysis notes that "non-serious situations" still fall under Class IIa, leaving minimal scope for the Class I default classification to apply in practice. This creates uncertainty precisely at the boundary where manufacturers need clarity most—determining whether their AI diagnostic or decision-support tool qualifies for the streamlined Class I pathway or requires the more burdensome Class IIa conformity assessment.

The revised rules retain only two risk dimensions from 2014 guidance—information significance and situation seriousness—rather than adopting a broader framework that would account for factors like autonomy, explainability, and workflow integration that are increasingly relevant for modern AI systems.

Clinical Evidence Trade-offs

The proposal does offer potential acceleration opportunities by broadening the definition of clinical data beyond peer-reviewed literature to include non-peer-reviewed scientific studies, though it remains unclear whether preprints qualify. The reforms also explicitly encourage New Approach Methodologies including in-silico testing for demonstrating safety and performance.

These expanded pathways could speed approval for certain AI device types, but they present a double-edged sword: faster market access balanced against questions about whether evidence robustness will remain adequate, particularly for high-risk AI applications in serious clinical situations.

The proposal also codifies a "Helsinki procedure" for resolving classification disputes through changes to MDR Article 4 and new Articles 4a, 51a, and 51b, while guidance documents from the EU's Artificial Intelligence Board and Medical Device Coordination Group attempt to clarify the MDR/IVDR and AI Act interplay.

Implications for Innovation and Access

For healthcare providers, the regulatory uncertainty may delay access to innovative diagnostic and therapeutic AI tools, particularly algorithmic decision support systems falling into ambiguous classification zones. Patients face continued delays in accessing AI-enabled medical devices in the EU market, while relaxed clinical evidence requirements raise questions about safety protections.

The European Parliament issued a Draft Report on the proposal by August 26, 2026, but the path forward remains unclear as stakeholders grapple with competing regulatory frameworks and timelines extending through 2028.

Sources

  1. The proposed reform of the Medical Device Regulation and its implications for AI-enabled medical devices in the EU — npj Digital Medicine