SaMD and Rule 11: Why the Reform Proposal Does Not Restore Class I

As of 4 August 2026. The Commission proposal COM(2025) 1023 final, submitted on 16 December 2025, provides for a complete rewrite of Rule 11 in Annex VIII MDR. In public discussion, the draft is widely presented as a return of Class I for medical device software (MDSW). A review of the wording does not support this reading.

Consequence for manufacturers: For any product whose conformity assessment must be completed before 2029, the reform proposal is strategically irrelevant. Rule 11 as currently in force and MDCG 2019-11 Rev. 1 remain authoritative.

1. Current legal framework

Classification of MDSW is governed by Annex VIII Chapter III Rule 11 MDR. The current version assigns software that provides information for diagnostic or therapeutic decisions to Class IIa, with escalation to IIb for serious deterioration or surgical intervention and to III for death or irreversible deterioration. The catch-all provision “all other software” remains in Class I.

The structural design flaw has been documented for years: the rule links exclusively to the severity of possible harm, not to risk as a combination of severity and probability. The Johner Institute therefore considers the Class I catch-all provision largely ineffective in practice.

The authoritative interpretive basis is MDCG 2019-11 Rev. 1, published on 17 June 2025. The revision is interpretive, not structural, but practically relevant in several respects: first introduction of the term Medical Device Artificial Intelligence (MDAI) as a distinct subset of MDSW, extension of Rule 11 to prognostic and predictive software, a substantially expanded section on modular software, and references to the European Health Data Space (EHDS). Technical documentation based on the 2019 version should be reviewed accordingly.

2. The Commission proposal: wording

COM(2025) 1023 final replaces paragraph 6.3 of Annex VIII in its entirety. The draft text classifies software intended to generate an output conferring clinical benefit and used for diagnosis, treatment, prevention, monitoring, prediction, prognosis, compensation or alleviation of a disease or condition as Class I, unless one of three exceptions applies: critical situation (Class III), serious situation or driving clinical management in a critical situation (Class IIb), or non-serious situation or informing clinical management (Class IIa). The terminology is clearly drawn from the IMDRF framework.

3. Analysis: where Class I actually remains in the draft text

The proposal reverses the logic: Class I is the starting point, with escalation via three exceptions. What matters is the scope of those exceptions.

Comparison with the IMDRF framework reveals the core problem: the third indent captures every “non-serious situation” and assigns it to Class IIa regardless of the significance of the information. The lowest category of the IMDRF framework has no equivalent in the draft text. Class I does not appear in the proposal’s matrix. A scope for Class I remains only outside the matrix, where the output confers clinical benefit but is not assigned to a specific clinical situation. BVMed therefore calls for Rule 11 to be drafted more precisely so that products can actually be placed in Class I.

4. Notified Body capacity

The widely cited figure of around 15,000 pending applications derives from the European Commission’s Notified Bodies Survey (data as of 31 October 2025, published March 2026). Cumulative MDR applications: 33,175; certificates issued: 17,549; arithmetic difference: 15,626. An independent survey by the Team-NB association confirms the order of magnitude from a second source.

Methodological caveat: The difference is not a queue size. It includes rejected, withdrawn and multiply counted applications as well as procedures leading to a single certificate covering multiple products. It represents an upper bound, not a stock figure. The substantiated finding is that the issuance rate falls significantly behind application volume. From 2026, recertification adds a further capacity constraint: the first wave of certificates issued in 2021 and 2022 is expiring.

5. Flanking legislation

Implementing Regulation (EU) 2026/977: Independently of the revision proposal, the Commission has adopted uniform quality management and procedural requirements for Notified Bodies. Core provisions include maximum deadlines of 30 days for application review and contract conclusion, 120 days for QMS audit and 90 days for product verification. Entry into force: 25 May 2026. Core application date: 25 February 2027. For ongoing procedures and agreements concluded before 25 February 2027, the existing framework continues to apply. From 2028, the reporting obligation creates for the first time a publicly comparable performance data basis across all Notified Bodies.

Relationship to the AI Act: The Commission proposal envisaged moving MDR and IVDR from Section A to Section B of Annex I to the AI Act. This shift was not adopted. The Council and Parliament agreed on 7 May 2026, in the context of the AI Omnibus, to keep MDR and IVDR in Section A. AI-enabled medical devices therefore remain subject to the full set of obligations for high-risk AI systems. As a compromise, the Commission is to be empowered to limit the applicability of the AI Act by implementing act where MDR or IVDR ensure an equivalent level of protection.

6. Legislative status and timeline

The reform course is politically contested. The TÜV Association, together with other Notified Bodies, warned Parliament and Council in an open letter in July 2026 against the Commission’s plans. On the manufacturer side, BVMed, Spectaris, VDGH, MedTech Europe and Pharma Deutschland are submitting amendments, including a sharpening of Rule 11. The rule text that emerges from trilogue will likely differ from the December draft.

7. Recommendations for action

Conclusion

The Commission proposal addresses a structural design flaw in Rule 11 that has been documented for years and draws for the first time on the IMDRF framework. That is the right regulatory direction. The current wording does not resolve the problem: because the third indent assigns the “non-serious situation” entirely to Class IIa, Class I has no place in the classification matrix. Key terms are undefined, and the conflation of patient vulnerability and severity of harm carries the severity-based design flaw into the new text.

By realistic estimate, binding application is still three to four years away. During that period, Rule 11 as currently in force and MDCG 2019-11 Rev. 1 remain the sole authoritative framework. The reform proposal is therefore not a planning parameter but an object of observation. Its practical value today lies solely in the further emphasis it places on a precisely formulated intended purpose and well-documented classification reasoning. Both carry weight under current law and will remain the starting point for classification decisions under any conceivable final version of the rule text.

Sources: (1) European Commission, COM(2025) 1023 final, 16.12.2025. (2) European Commission, MDCG 2019-11 Rev. 1, 17.06.2025. (3) European Commission, Notified Bodies Survey, data as of 31.10.2025. (4) Team-NB, Survey on MDR and IVDR implementation, end 2025. (5) IR (EU) 2026/977, OJ 05.05.2026. (6) Council of the EU, AI Omnibus, 07.05.2026. (7) Council of the EU, EPSCO, 16.06.2026. (8) VDE, Rule 11 recast, 13.01.2026. (9) Johner Institute, MDR Classification Rule 11, 21.01.2026. (10) IMDRF/SaMD WG/N12FINAL:2014. (11) MedTech Europe, MDR/IVDR Revision, 05.05.2026. (12) BVMed, MDR revision position, May/June 2026. (13) MDCG 2025-6, FAQ MDR/IVDR and AI Act, June 2025.