High-Risk AI Systems: When Does Medical Device AI Qualify?

The AI Act, Regulation (EU) 2024/1689, was adopted on 13 June 2024 and applies generally from 2 August 2026. Regulation (EU) 2026/1744, the Digital Omnibus on AI, was adopted on 8 July 2026 and entered into force on 27 July 2026. It moved the high-risk requirements to 2 December 2027 for Annex III systems and to 2 August 2028 for systems that are high-risk under Article 6(1) and Annex I. For Article 6(1) systems, the original date was 2 August 2027.

The date has moved, but whether medical device AI is high-risk at all still depends on its MDR or IVDR class. This article sets out the classification test, how MDCG 2025-6 maps device classes onto it, what follows for providers, and the dates that now apply.

Article 6(1): two cumulative conditions

Under Article 6(1), an AI system is high-risk when both of the following conditions are met:

1. the AI system is intended to be used as a safety component of a product, or is itself a product, covered by the Union harmonisation legislation listed in Annex I; and

2. that product is required to undergo a third-party conformity assessment under that legislation before it is placed on the market or put into service.

The MDR and IVDR are listed in Annex I, Section A. AI software that is itself a medical device or IVD, or that is a safety component of one, therefore meets the first condition. The second condition depends on whether a notified body is involved in the conformity assessment.

Regulation (EU) 2026/1744 added three new paragraphs that clarify both conditions:

• Article 6(1a): AI systems used solely for non-safety aspects, such as user assistance, performance optimisation, service efficiency, automation, convenience or quality control, do not qualify as safety components.

• Article 6(1b): AI systems whose failure or malfunctioning would endanger health and safety do qualify as safety components.

• Article 6(1c): a product that needs third-party assessment solely because of risks other than health and safety does not meet the second condition.

In practice, the AI Act answer follows the device file. Intended purpose comes first, then the MDR or IVDR class, then the conformity assessment route, and only then the Article 6(1) conclusion.

How MDCG 2025-6 maps MDR and IVDR classes

MDCG 2025-6, the MDCG and AI Board FAQ on the interplay between the MDR, IVDR and AI Act (June 2025), translates the second condition into device classes:

Source: MDCG 2025-6, Table 1.

Two points follow. First, high-risk status under the AI Act does not change the MDR or IVDR risk class (MDCG 2025-6, Q4). A class IIa device stays class IIa. Second, Rule 11 of the MDR places software that provides information for diagnostic or therapeutic decisions in class IIa or higher, so AI-enabled SaMD with that purpose will meet both conditions.

MDCG 2025-6 was published before the Digital Omnibus and still refers to 2 August 2027 as the date from which Article 6(1) applies. The mapping is unaffected, but the date is now 2 August 2028.

Annex III and the Article 6(3) derogation

Article 6(2) adds a second, separate route: AI systems used in the areas listed in Annex III are also high-risk. Annex III covers biometrics, critical infrastructure, education, employment, access to essential private and public services, law enforcement, migration and border control, and the administration of justice and democratic processes. Two of its entries can touch healthcare: AI used to evaluate eligibility for essential public assistance benefits and services, including healthcare services, and emergency healthcare patient triage systems.

Article 6(3) allows an Annex III system not to be considered high-risk when it does not pose a significant risk of harm, for example because it only:

• performs a narrow procedural task;

• improves the result of a previously completed human activity;

• detects decision-making patterns or deviations without replacing or influencing the human assessment without proper human review; or

• performs a preparatory task to an assessment.

An Annex III system that profiles natural persons is always high-risk. Under Article 6(4), a provider that relies on the derogation must document its assessment before placing the system on the market and is subject to registration under Article 49(2).

In our view, the derogation qualifies only the Annex III route. On our reading of the text, it offers no way out for a device that is already high-risk under Article 6(1).

What high-risk status requires

A high-risk AI system must meet the requirements in Chapter III, Section 2:

Regulation (EU) 2026/1744 amended Article 11(1): SMEs, including start-ups, and small mid-cap companies may provide the technical documentation in a simplified form established by the Commission, and notified bodies must accept that form.

For medical devices, these requirements are assessed within the MDR or IVDR conformity assessment rather than in a separate procedure. Our article on EU AI Act requirements for AI-enabled medical devices covers that interplay, and our overview of AI/ML standards for medical devices covers the standards being developed to support it.

Provider obligations and the QMS

Article 16, points (a) to (l), lists the provider's obligations. They include meeting the Section 2 requirements, a quality management system under Article 17, documentation keeping, log retention, conformity assessment, an EU declaration of conformity, CE marking, registration, corrective action and accessibility.

Article 17(2) already required the QMS to be proportionate to the size of the provider's organisation. Regulation (EU) 2026/1744 adds an explicit reference to SMEs, including start-ups, and small mid-caps. MDCG 2025-6 (Q6) confirms that manufacturers may include the AI Act QMS elements in their existing MDR or IVDR quality management system. One system, with two legal bases, is the intended outcome.

New Article 2(13) may reduce the overlap further. For Article 6(1) systems, specific requirements in Articles 9 to 15 and 17 to 25 may be limited where the Annex I, Section A legislation provides an equivalent or higher level of protection, and the limitation does not reduce the overall level of protection under the AI Act. The Commission must adopt delegated acts specifying this by 2 August 2027. Until those acts exist, their content is unknown, and the safer plan is to build evidence against the requirements in full.

The dates that now apply

Sources: Article 113, as amended; Council press release, 29 June 2026.

For medical device AI that is high-risk under Article 6(1), the relevant date is 2 August 2028.

A hypothetical example

Hypothetical example: a Dutch start-up develops software that uses a trained model to flag findings on medical images for review by a clinician.

The first step is qualification and classification under the MDR. If the software qualifies as a medical device and its class requires a notified body, MDCG 2025-6 Table 1 indicates that the Article 6(1) conditions are fulfilled. Its failure could also endanger health, so it falls within Article 6(1b) rather than the non-safety carve-out in Article 6(1a).

The start-up then builds the Article 9 to 15 evidence, including data governance, logging, human oversight design, accuracy and cybersecurity, into the technical documentation it already maintains under the MDR. Our article on SaMD technical documentation covers that structure. It extends its existing QMS to cover the Article 17 elements and plans against 2 August 2028. Because data governance and logging evidence has to cover the releases that lead up to that date, the work starts now rather than in 2028.

Conclusion

In our view, for medical device AI the MDR or IVDR classification largely answers the AI Act question. Notified body involvement is what triggers Article 6(1), and MDCG 2025-6 maps that link class by class. The Article 6(3) derogation addresses Annex III systems, not Article 6(1) devices.

2 August 2028 is a planning date, not a reason to wait. Data governance, logging and change-control evidence need to exist across releases, and the AI Act elements belong in the existing QMS and technical documentation. The delegated acts under Article 2(13) may narrow some requirements later, but they will not remove the classification question.

If you need support assessing whether your device AI is high-risk and planning the evidence, our regulatory strategy team can help, or contact MedQAIR.

Frequently asked questions

Is every AI medical device a high-risk AI system?

No. Article 6(1) requires both an Annex I product and a third-party conformity assessment. Under MDCG 2025-6, devices without notified body involvement, such as non-sterile, non-measuring MDR class I devices, non-sterile IVDR class A devices and in-house devices, do not meet the second condition.

Does high-risk status change my MDR or IVDR class?

No. MDCG 2025-6 (Q4) states that the AI Act does not affect the risk classification of medical devices or IVDs.

Can I use Article 6(3) to avoid high-risk status?

Article 6(3) applies to Annex III systems. On our reading of the text, it does not apply to a device that is high-risk under Article 6(1). Annex III systems that profile natural persons are always high-risk.

When do the high-risk requirements apply?

Following Regulation (EU) 2026/1744, from 2 December 2027 for Annex III systems and from 2 August 2028 for Article 6(1) systems, the route through which medical devices and IVDs become high-risk.

Can the AI Act QMS be combined with my MDR or IVDR QMS?

Yes. MDCG 2025-6 confirms that the AI Act QMS elements may be included in the existing MDR or IVDR quality management system.