EU Commission Drafts Guidelines on Classifying High-Risk AI Systems 

23 Jul 2026

Introduction: A New Layer of Compliance 

Artificial Intelligence (AI) is reshaping industries across Europe, but nowhere is the impact more profound than in healthcare and medical devices. With the EU Artificial Intelligence Act (AI Act) set to take effect in December 2027, the European Commission has now published draft guidelines to clarify how companies should classify their AI systems as high-risk. You can access the full European Commission draft guidelines online. 

These guidelines are not just technical notes — they represent a strategic roadmap for manufacturers, regulators, and innovators. For medical devices, the AI Act does not replace existing frameworks like the Medical Device Regulation (MDR) or the In Vitro Diagnostic Regulation (IVDR). Instead, it layers additional obligations on top, creating a dual compliance challenge that companies must navigate carefully. 

The Storyline: From Regulation to Reality 

The AI Act was born out of Europe’s ambition to create “human-centric and trustworthy AI.” But as the law moves closer to implementation, questions arise: Which systems are truly high-risk? How do companies avoid double regulation? And what does this mean for innovation in healthcare? 

The Commission’s draft guidelines attempt to answer these questions by: 

• Defining classification principles under Article 6. 

• Providing examples of high-risk systems across industries. 

• Clarifying how AI obligations interact with MDR and IVDR. 

This storyline is not just about compliance — it is about balancing innovation with safety, ensuring that AI-driven medical devices can reach patients without being buried under regulatory complexity. 

Key Principles of Classification 

The guidelines emphasize that the intended purpose of an AI system is central to determining risk. 

• Article 6(1) AI Act: AI systems are high-risk if they act as safety components of products or require third-party conformity assessments. 

• Annex I Guidance: Defines “safety functions” such as monitoring abnormal behavior or detecting worn-out safety parts. 

• Annex III Guidance: Lists AI systems with potential impact on health, safety, and fundamental rights — subject to strict compliance obligations. 

The Commission notes that these examples are not exhaustive and will evolve over time. 

Impact on Medical Devices 

AI in medical devices already falls under MDR/IVDR. The AI Act adds a second compliance layer for high-risk systems. 

Examples of High-Risk Medical AI: 

• Pure AI software: Algorithms that analyze scans to diagnose diseases. 

• AI integrated into hardware: Infusion pumps or surgical robots where AI regulates dosing or controls safety functions. 

By contrast, Class I self-certified devices remain outside the high-risk category. 

This distinction matters because it determines whether a device requires independent conformity assessment and registration under the AI Act. 

Avoiding Double Regulation 

The Commission acknowledges the risk of dual regulation and is working on reforms: 

• Digital Omnibus on AI: 

• Postpones high-risk obligations until August 2028. 

• Allows exemptions where MDR/IVDR already mandate equivalent standards. 

• MDR/IVDR Revision 2025: 

• Proposes integrating AI requirements directly into device rules. 

• Mirrors how AI is regulated in automotive and robotics sectors. 

These reforms aim to streamline compliance and prevent companies from duplicating efforts across frameworks. 

Compliance Timeline 

• 2 December 2027 → Stand-alone high-risk AI systems. 

• 2 August 2028 → High-risk AI systems embedded in products. 

This phased timeline gives companies more time to adapt, but also signals that compliance preparation must begin now. 

Broader Coverage of the AI Act 

The guidelines extend beyond healthcare, covering: 

• Safety components of machinery, toys, lifts, and protective systems. 

• Products in automotive, aviation, and radio equipment. 

• Medical devices and IVDs — the most sensitive category due to direct impact on human health. 

This breadth shows the EU’s intent to create a unified AI safety framework across industries. 

Compliance Obligations for High-Risk AI 

High-risk AI systems must meet strict obligations, including: 

• Independent conformity assessments. 

• Risk management systems. 

• Registration requirements. 

• Evidence of data governance, robustness, and human oversight. 

For medical device manufacturers, this means aligning AI documentation with MDR/IVDR processes to avoid duplication. 

Industry Perspective 

The sensible approach is to evidence AI-specific points — such as data governance and human oversight — through MDR/IVDR documentation rather than duplicating them under the AI Act. 

This approach reflects a practical balance: medical AI remains regulated, but companies gain more time and clarity, with the focus shifting to device rules. 

Conclusion: Trustworthy AI in Healthcare 

The EU’s draft guidelines mark a pivotal step in defining high-risk AI systems under the AI Act. For medical devices, the message is clear: 

• Compliance will be layered on top of MDR/IVDR. 

• High-risk classification depends on intended purpose and conformity assessment. 

• Reforms are underway to prevent dual regulation. 

For innovators, this is not just about meeting regulatory deadlines — it is about building trustworthy, safe, and human-centric AI that can transform healthcare. 

The storyline is one of balance: between innovation and regulation, between safety and speed, and between European ambition and global competitiveness. 

Empowering Engineering Excellence with TS Quality & Engineering 

At TS Quality & Engineering, we view these guidelines not as a compliance burden, but as a framework for engineering excellence. We specialize in transforming complex regulatory requirements into robust, auditable quality management systems. Whether you are navigating Annex I requirements or seeking to integrate AI governance into your standard development lifecycle, our team provides the engineering rigour needed to turn compliance into a competitive advantage. By bridging the gap between rigorous technical standards and regulatory compliance, we empower medical device innovators to deliver safe, high-quality, and compliant AI solutions to market faster. Let us help you future-proof your medical AI. Contact TS Quality & Engineering to start your compliance transformation.