Skip to main content
  • Regulatory Update

Risk Categorization Per the European AI Act

Our Emergo by UL experts explain the importance of risk categorization in complying with the European Artificial Intelligence Act and Medical Device Regulation (2017/745 MDR) and In-Vitro Diagnostic Regulation (2017/746 IVDR).

AI and MI enabled medical devices

April 1, 2025

By Sade Sobande and Evangeline Loh

In this blog, we discuss how the European Artificial Intelligence Act (AI Act) is now part of the new legislative framework.

Our webinar on the European Artificial Intelligence Act Regulation 2024/1689 emphasized the importance of risk categorization per the Medical Device Regulation (MDR, 2017/745) and In-Vitro Diagnostic Regulation (2017/746 IVDR) and the AI Act.

The AI Act is now part of the new legislative framework

The AI Act is a new part of the new legislative framework and also applies to medical devices and IVDs.

One of the fundamental tenets is that the medical device and IVD are subject to EU harmonized legislation: the MDR and IVDR are delineated in Annex I of the AI Act.

Categorization as high-risk AI system for medical devices or IVDs

Article 6 of the AI Act defines the criteria for an AI system to be classified as high-risk. According to paragraph 1(b), an AI system is considered high-risk if it is intended to be used as a safety component of the medical device or IVD, or the AI system is itself a product where the risk of the medical device or IVD requires Notified Body review per the MDR or IVDR, respectively. Therefore, an AI-enabled medical device or IVD, which requires Notified Body involvement, would likely be categorized as a high-risk AI system. Exceptions may exist in the rare situation where a higher-risk medical device or IVD employs AI that is not a safety component of the device.

As we’ve advised, the classification needs to be clearly documented and substantiated within the technical documentation. It is expected that a significant number of AI-enabled medical devices and IVDs will fall under the high-risk category. To facilitate regulatory alignment, the use of structured dialogues may be one mechanism to engage a Notified Body in discussion related to this and support clarity and consistency in the classification process.

Categorization as AI Act limited risk medical devices or IVDs

Based on Article 6, if a medical device or IVD is classified as Class I self-certified or Class A non-sterile, respectively, it would not be categorized as a high-risk AI system, as it would never fulfill Article 6(1)(b).

Regulatory requirements for high-risk AI systems

The majority of the AI Act revolves around describing requirements for high-risk AI systems (Chapter III, Articles 6-49).

Our previous regulatory updates have extensively discussed the requirements for high-risk AI systems as this has been a popular topic: history, scope, classification, conformity assessment and operator’s obligations of AI Act, AI Act best practices, enforcement and oversight, and AI regulatory sandboxes and real-world testing, penalties and timelines and deadlines and requirements for high-risk AI systems, the similarity to the EU MDR/IVDR, additional regulatory requirements and the route to market.

What are the regulatory requirements when the medical device or IVD isn’t also a high-risk AI system?

Interestingly, a Declaration of Conformity (DoC) is not required (Article 16(g)). The regulatory requirements for limited-risk medical devices and IVDs are outlined in Chapter IV, transparency obligations for providers and deployers of certain AI systems, Article 50. These requirements revolve around informing users that they are engaging with AI and making users aware that the AI system is generating content.

Concluding remarks

While most articles describe high-risk AI systems, Class I self-certified medical devices and Class A IVDs would never be categorized as a high-risk AI system.  

Based on Article 6, the majority of AI-enabled medical devices and IVDs would likely also be categorized as high-risk AI systems. If manufacturers have categorized their higher risk medical device or IVD as not a high-risk AI system, manufacturers must document a clear rationale and may try to engage their Notified Bodies in structured dialogues related to the whether the medical device or IVD has been appropriately categorized.

As always, we at Emergo by UL are available to assist you with any questions you have about this process.

X

Request more information from our specialists

Thanks for your interest in our products and services. Let's collect some information so we can connect you with the right person.

Please wait…