Smarter regulation, not deregulation: why the Parliament’s position on AI in medtech points in the right direction

AI is already in the operating theatre, the imaging suite and the diagnostic laboratory: reading scans, flagging anomalies, predicting patient deterioration before clinical signs appear. The question for European policymakers has never been whether to regulate it. The question is how to regulate it well. 

Last month, the European Parliament adopted its negotiating position on the Digital Omnibus on AI: a serious piece of regulatory architecture that will allow Europe to keep the door open to innovation in Digital Health. Trilogue has now formally begun, with technical and political meetings scheduled for the upcoming weeks. The clock is not abstract: if the Digital Omnibus is not adopted before 2 August 2026, the original high-risk AI obligations apply as currently written, before the standards and implementing measures needed to make them workable are in place. MedTech Europe has welcomed the delay in application dates as a pragmatic necessity; what must not be delayed is the substance. 

What the European Parliament decided 

The Parliament chose not to delete or weaken the AI Act’s high-risk requirements as they apply to medical devices. Human oversight, transparency, robustness, data governance and logging: these safeguards remain intact. High-risk AI requirements for medical devices are to be addressed within the existing Medical Devices Regulation (MDR) and In Vitro Diagnostics Regulation (IVDR) conformity assessment process, by the specialised notified bodies already responsible for that work. 

MedTech Europe welcomes this direction. There was a real risk that the Digital Omnibus process could strip away AI-specific obligations under the banner of simplification. Simplification is a legitimate and necessary goal, and the ongoing MDR/IVDR revision is the right vehicle for it. The goal is a regulatory system that is both lean and safe. Those two objectives are compatible, provided the right things are simplified. The Parliament understood that distinction. The task now is for Member States and the European Commission to preserve it. 

Why this architecture makes sense 

Keeping the AI Act and the MDR/IVDR as two overlapping sets of rules would fragment the framework, increase administrative complexity, and make it harder to assess AI risks in their full clinical context. Consider two examples from our membership. A company developing AI-assisted pathology software for cancer detection must already demonstrate clinical validity, post-market surveillance, and lifecycle safety management under the MDR: a separate conformity pathway for the AI component would duplicate that process without adding a single meaningful patient safety check. A manufacturer of AI-enabled continuous glucose monitoring systems follows the same logic. The risk to the patient is inseparable from the device; therefore, from a patient safety perspective, it is important to manage AI medical technology risks in the device framework. 

Moving medical devices from Annex I Section A to Section B of the AI Act doesn’t mean reducing oversight, control or safety verifications, but ensures they are applied coherently within the MDR/IVDR framework. AI-enabled medical technologies would remain fully subject to the relevant requirements in the AI Act, including risk management, data governance, transparency and human oversight. The difference is that these requirements would be assessed through the IVDR/MDR system already designed to evaluate them in their full clinical context. 

The MDR and IVDR already require manufacturers to address risks specific to their technology, including software and algorithmic systems. Annex I is the mechanism designed for precisely this purpose. Developing specific AI requirements there, explicitly reflecting Chapter III, Sections 1, 2 and 3 of the AI Act, is not a workaround. It is the framework working as intended. 

The transition risk is real 

Some critics have raised a legitimate concern: embedding AI safeguards via future delegated acts and standards creates a period of legal uncertainty during which neither regime is clearly applicable. AI-enabled devices are in clinical use today. Patients cannot wait for a multi-year standard-setting process to conclude before the rules are clear. 

We share that concern. The Commission now has both the mandate and the opportunity to move quickly. Laying down specific AI requirements in Annex I, with clear application dates, would give manufacturers and notified bodies the certainty they need from the outset. Speed and specificity together are what transform a sound architectural decision into workable rules on the ground. 

Notified bodies, too, must be equipped for this role. Assessing AI-enabled devices requires specialist competencies not uniformly present across the network today. MedTech Europe supports targeted investment in notified body capacity and AI-specific designation criteria as essential conditions for effective implementation. 

What is at stake 

Earlier cancer detection. More accurate glucose monitoring. Personalised surgical planning. The pipeline is substantial and the potential to improve outcomes, reduce costs and extend healthcare’s reach to underserved populations is real. None of it is realised in a regulatory vacuum. 

The Digital Omnibus package covers significant ground beyond the AI Act, including the GDPR and the Data Act, on all of which MedTech Europe has submitted detailed positions. This article focuses on the AI Act and MDR/IVDR intersection, where the stakes for patients and for the sector are most immediate. 

Manufacturers need legal certainty and an efficient system. Notified bodies need clear criteria. Patients need assurance that the clinical AI tools used on them today meet a robust set of requirements which are assessed across their lifecycle. The Parliament’s negotiating position creates the conditions for all three. Along with the MDR and IVDR revision, the Digital Omnibus is the perfect opportunity to deliver them.