Short answer

If your medical software uses machine learning to generate clinical outputs and needs a Notified Body under the MDR (class IIa, IIb or III), it is a high-risk AI system under Article 6(1) of the AI Act; there is no separate AI risk classification. After the Digital Omnibus, Regulation (EU) 2026/1744 (in force since 27 July 2026), the high-risk requirements apply to AI medical devices from 2 August 2028 instead of 2 August 2027. Much of the work builds on your MDR documentation. The main new areas are data governance (Article 10), transparency in the instructions for use (Article 13) and human oversight by design (Article 14). EN 18286 is the first harmonised standard for the AI Act QMS.

1. Does the EU AI Act apply to your software?

The EU AI Act (Regulation (EU) 2024/1689) defines an AI system in Article 3(1) as a machine-based system designed to operate with varying levels of autonomy, which may exhibit adaptiveness after deployment and which, for explicit or implicit objectives, infers from the input it receives how to generate outputs such as predictions, content, recommendations or decisions. If your SaMD uses a machine-learning model to produce clinical outputs, it will very likely meet this definition. Software that only executes rules defined solely by people is not covered (recital 12), so assess each component rather than the product label.

The risk class question resolves quickly for most medical software companies. Under Article 6(1), an AI system is high-risk if it is itself a product, or a safety component of a product, covered by the legislation listed in Annex I — Section A of which includes the MDR and IVDR — and that product has to undergo third-party conformity assessment. There is no separate AI risk classification exercise: whether your device needs a Notified Body under the MDR determines its AI Act status.

The practical rule

Class IIa, IIb or III under EU MDR (Notified Body conformity assessment) = high-risk AI under Article 6(1). Class I devices that do not need a Notified Body are generally outside this route — unless the AI is a safety component of a product that does need third-party assessment, or the use falls under Annex III (for example, emergency healthcare patient triage). MDCG 2025-6 (June 2025) sets out the same logic for medical device AI.

If your software is not yet classified under EU MDR, that classification question comes first. See our guide on EU MDR Rule 11 and SaMD classification before proceeding.

2. The dates — original, current, and what already applies

This is the section that has changed most since this guide was first published. Regulation (EU) 2026/1744 — the "Digital Omnibus on AI" — was signed on 8 July 2026, published in the Official Journal on 24 July 2026 and entered into force on 27 July 2026. It replaced the original high-risk application dates in Article 113 of the AI Act with later, fixed dates. Medical devices — AI systems that are themselves a medical device, or a safety component of one — fall under Article 6(1) and Annex I as product-embedded high-risk AI, not Annex III. Here is where things stand:

2 August 2027
(original date)
Replaced. This was the original date for the high-risk obligations under Article 6(1) and Annex I, including AI medical devices that need a Notified Body. Regulation (EU) 2026/1744 moved it by one year.
2 August 2028
(date now in force)
The high-risk requirements (Chapter III, Sections 1–3) apply to AI SaMD under Article 6(1) and Annex I. For devices already placed on the market or put into service before that date, Article 111(2) contains a transitional rule: the high-risk obligations reach them only if they are subsequently subject to significant changes in their design. How this applies to a continuously released software product is worth discussing early with your Notified Body.
2 December 2027 Annex III (stand-alone) high-risk AI — recruitment, credit scoring, education, law enforcement and similar uses, including two with a healthcare link: public-authority decisions on eligibility for healthcare services and emergency healthcare patient triage systems. Most SaMD sits under Annex I instead; this date is listed because summaries often conflate the two.
What already applies today

The postponement only concerns the high-risk requirements. Under Article 113, other parts of the AI Act already apply: the AI literacy obligation (Article 4) and the prohibited practices since 2 February 2025, the obligations for providers of general-purpose AI models since 2 August 2025, and the Article 50 transparency obligations — for example, informing people that they are interacting with an AI system — since 2 August 2026. Regulation (EU) 2026/1744 reworded Article 4: providers and deployers must take measures to support the AI literacy of their staff, without having to guarantee a specific level for each individual. If you build AI SaMD, AI literacy measures for your team are already due; Article 50 matters if your product includes, for example, a patient-facing chatbot.

3. What your existing MDR work already covers

The good news: substantial portions of the AI Act's high-risk requirements map onto EU MDR and ISO 13485 obligations you are likely already meeting. MDCG 2025-6 (June 2025, endorsed by the MDCG and the AI Board) explains how the two frameworks are meant to work together — for example, a single set of technical documentation for high-risk AI medical devices (AI Act Article 11(2)) and the option to integrate AI Act testing, reporting and documentation into existing MDR procedures (Article 8(2)). Note that MDCG 2025-6 predates Regulation (EU) 2026/1744, so the 2 August 2027 date it quotes has since been replaced.

Requirement area MDR / ISO 13485 coverage AI Act status
Quality Management System ISO 13485 QMS; AI Act Article 17(3) allows the AI-specific QMS elements to be included in it. EN 18286:2026 is the harmonised standard for this QMS Partial — extend for AI processes
Risk management ISO 14971 process; AI Act Article 9 risk management can be combined with it (Article 9(10)) Partial — add AI-specific risks
Technical documentation MDR Annex II/III technical documentation; one single set covers both (Article 11(2)) Partial — add data governance sections
Conformity assessment Notified Body MDR assessment — AI Act requirements are assessed within it (Article 43(3)) Largely covered
Post-market surveillance MDR PMS plan and PMCF Partial — add AI monitoring metrics
Data governance documentation Not explicitly required under MDR New — Article 10 requirement
Transparency / IFU for AI outputs IFU requirements exist but less specific New — Article 13
Human oversight by design Not explicitly required under MDR New — Article 14 design requirement
EU database for high-risk AI Device registration in EUDAMED under MDR Only if an Annex III use applies (Article 49)

4. What is genuinely new — the three main gaps

Despite the overlaps, the AI Act introduces three requirement areas with no direct MDR equivalent. These are where companies with solid MDR files still have meaningful work to do.

Gap 1 — Data governance (Article 10)

The AI Act requires documented data governance practices covering the ML pipeline: data collection, labelling, cleaning and preparation, and — critically — the quality of training, validation and test datasets. Article 10 requires them to be relevant, sufficiently representative and, to the best extent possible, free of errors and complete in view of the intended purpose, taking into account the specific setting in which the system will be used. It also requires an examination of possible biases and measures to detect, prevent and mitigate them. For a medical device this usually means showing that the datasets reflect the intended patient population — for example by age, sex, clinical setting and relevant comorbidities — and documenting how bias was examined and mitigated.

In practice, this means retroactively documenting decisions made during model development. In our experience, this is the most time-consuming AI Act task — dataset provenance records rarely exist in the form the AI Act requires.

Gap 2 — Transparency and instructions for use (Article 13)

High-risk AI systems must be transparent enough to enable deployers to interpret the system's output and use it appropriately. Article 13(3) lists what the instructions for use must contain, including the level of accuracy and its metrics, known or foreseeable circumstances that may lead to risks, performance for specific persons or groups where appropriate, any pre-determined changes, and the human oversight measures. A generic "for physician review only" disclaimer does not cover this. These requirements are more specific about AI performance and limitations than the general MDR labelling requirements.

Gap 3 — Human oversight by design (Article 14)

This is a design requirement, not a documentation requirement. Your product architecture must allow clinical users to stop, override, or disregard AI outputs at appropriate decision points. If your current system makes it difficult in practice to override an AI recommendation — whether by UX design or workflow integration — redesign may be necessary before 2 August 2028.

Not a gap for most AI SaMD — the EU database for high-risk AI

Earlier versions of this guide listed registration in the EU database for high-risk AI systems as a fourth gap. That was too broad. Article 49(1) requires registration for high-risk AI systems listed in Annex III; AI that is high-risk only because it is a medical device under Article 6(1) is not on that list, and the device itself is registered in EUDAMED under the MDR. If part of your product serves an Annex III use — for example an emergency healthcare patient triage function — check Article 49 for that part.

Your Notified Body and the AI Act

Your technical documentation will need to satisfy both MDR Annex II/III and AI Act Annex IV, as a single set (Article 11(2)). Under Article 43(3), AI medical devices follow the MDR conformity assessment procedure, and the AI Act requirements are assessed as part of it; Regulation (EU) 2026/1744 requires notified bodies that assess such systems to apply for designation under the AI Act by 28 January 2028. Ask your NB directly how it is approaching AI Act requirements in its MDR assessments — approaches still differ between bodies and affect how you structure your documentation.

Not sure what your AI Act gap looks like? I can review your MDR technical documentation and list what the AI Act will add for your device before 2 August 2028.
Request a free call →

5. The Digital Omnibus — what was adopted

Earlier versions of this guide described the Digital Omnibus first as an open question and then as a provisional deal. It is now law. The Commission proposed it on 19 November 2025; the Council agreed its position on 13 March 2026; Council and Parliament negotiators reached a provisional agreement on 7 May 2026; Parliament adopted its position on 16 June 2026 and the Council gave its final approval on 29 June 2026. Regulation (EU) 2026/1744 was signed on 8 July 2026, published in the Official Journal on 24 July 2026 and entered into force on 27 July 2026.

The key outcomes relevant to medical AI:

  • Annex I product-embedded high-risk AI (including medical devices) — obligations postponed by one year, from 2 August 2027 to 2 August 2028.
  • Annex III stand-alone high-risk AI (recruitment, credit scoring, law enforcement, education, border control and similar uses) — obligations postponed by sixteen months, from 2 August 2026 to 2 December 2027.
  • Both dates are fixed in Article 113 itself.
  • A separate change: each Member State must have at least one national AI regulatory sandbox operational by 2 August 2027 (previously 2 August 2026).
  • A new mechanism (Article 2(13)) allows the Commission to limit, by delegated acts to be adopted by 2 August 2027, specific requirements (Articles 9–15 and 17–25) for Annex I Section A products — which include medical devices — where the sectoral legislation already provides an equivalent or higher level of protection, without reducing the overall level of protection. Which requirements, if any, will be limited for MDR/IVDR devices is not yet known.
  • Medical devices remain within Annex I Section A high-risk AI; the Digital Omnibus did not move them out. They simply get more time.
Where this stands — and the separate MDR/IVDR proposal

Regulation (EU) 2026/1744 is in force and directly applicable in all Member States; the dates above are law. Separately, the Commission's proposal to simplify the MDR and IVDR — COM(2025) 1023 of 16 December 2025 — would move the MDR and IVDR from Section A to Section B of Annex I to the AI Act, which would limit the AI Act's direct application to medical devices. This is a proposal only: as of October 2026 it is still being examined by the European Parliament and the Council. Do not build your plan on it until it is adopted.

The practical shift for most medical AI developers: the same underlying work — data governance documentation, transparency and IFU updates, human oversight by design, QMS extension — still needs to happen, but the runway has meaningfully lengthened. Instead of the original 2 August 2027 date, companies now have until 2 August 2028 to plan and execute it properly, in step with their ongoing MDR technical file work.

That said, "more time" is not "no urgency." The first harmonised standard is already available: EN 18286:2026, for the quality management system, was cited in the Official Journal on 14 October 2026. Standards for the other high-risk requirements, such as risk management, data and cybersecurity, are still being revised by CEN and CENELEC, and each gives a presumption of conformity only once its reference is published in the Official Journal. Companies seeking initial CE certification in the near term should build AI Act-aligned documentation into their technical file from the outset — retrofitting data governance documentation after the fact remains the most time-consuming part of this work regardless of which date applies.

6. Six-step action plan for 2028

If your SaMD uses ML and is Class IIa or above under MDR, these are the concrete steps to complete before 2 August 2028. With the extended timeline there is no need to compress this work, but there is also no reason to defer it entirely. Building it into your next MDR technical file revision cycle is the most efficient path.

  1. 1
    Inventory every AI component in your product Document each ML model: its purpose, training data source, version history, and the clinical outputs it generates. This is the foundation for AI Act Annex IV technical documentation and your ongoing monitoring plan. If you cannot articulate this clearly, your AI Act compliance work cannot begin.
  2. 2
    Gap-assess your technical file against AI Act Annex IV Compare your current MDR technical file section by section against the nine points of Annex IV. Data governance documentation (training data provenance, bias assessment methodology) and explainability approach are the most common gaps in files we review. A structured technical documentation audit surfaces these gaps systematically — worth doing now even with the extended deadline, since it folds naturally into your next MDR technical file update.
  3. 3
    Document your training datasets retroactively Reconstruct and document data collection methodology, labelling process, dataset composition by relevant demographic subgroups, and the bias identification and mitigation steps taken during development. This is typically the most time-consuming step for teams that did not document as they built.
  4. 4
    Update your PMS plan for continuous AI monitoring Add AI-specific performance metrics to your existing post-market surveillance plan: model accuracy drift by patient subgroup, out-of-distribution detection rates, and defined thresholds for triggering model review or withdrawal. This overlaps with MDR PMCF requirements and can be integrated into your existing PMS framework.
  5. 5
    Extend your ISO 13485 QMS to cover AI-specific processes Add at minimum: a procedure for managing training data versions, a bias assessment process, and a model change management procedure that determines when a post-deployment model change needs Notified Body approval under MDR — and, under the AI Act, whether it falls within the changes you pre-determined (Article 43(4)). Our guide to change control for AI SaMD under EU MDR covers this in detail. Extending an existing QMS for these AI processes is usually far less work than building one.
  6. 6
    Contact your Notified Body and track the open points Ask your NB directly how it is integrating AI Act requirements into its MDR assessments and whether it has applied for designation under the AI Act — approaches differ, and they affect your documentation strategy. In parallel, watch for the Commission's delegated acts under Article 2(13), the citation of further harmonised standards (the first, EN 18286 for the QMS, was cited on 14 October 2026), and the progress of the MDR/IVDR proposal COM(2025) 1023.

Companies with solid MDR technical files and ISO 13485 QMS in place are significantly better positioned than those starting from scratch. The gap between "MDR compliant" and "AI Act compliant" is real — but it is bridgeable when approached systematically. With the high-risk requirements now applying from 2 August 2028, the sensible move is to fold this work into your regular MDR documentation cycle rather than treat it as a separate fire drill.

Official sources & references