EU AI Act Annex I High-Risk Rules Move to 2 August 2028

Regulation (EU) 2026/1744 of 8 July 2026 — the Digital Omnibus on AI — amends the Artificial Intelligence Act (Regulation (EU) 2024/1689). It was published in the Official Journal on 24 July 2026. Article 4 says it enters into force on the third day following publication. That makes the entry-into-force date 27 July 2026 (Pure Global calendar calculation). The text has EEA relevance.

For medical-device and IVD manufacturers the operational split is this: Article 50's general application date has passed. The high-risk obligations in Chapter III, Sections 1, 2 and 3 (except Article 6(5)) now apply later than the AI Act originally provided, and later still for products covered by Annex I — the list that includes Regulations (EU) 2017/745 and 2017/746. New Article 111(4) provides one narrower transition for Article 50(2) marking, explained below.

Finland’s Fimea restated that split for the device sector on 25 August 2026. The controlling wording is the Official Journal text, not the national notice.

This article covers the application dates that change device clocks, the Article 50 marking deadline for systems already on the market, and the amended “safety component” test. It does not catalogue every Omnibus amendment.

What applied when, before and after

Under the unamended Article 113 of Regulation (EU) 2024/1689:

• the Act applied from 2 August 2026, except as listed below;
• Chapters I and II applied from 2 February 2025;
• Chapter III Section 4, Chapters V, VII and XII, and Article 78 applied from 2 August 2025 (except Article 101);
• Article 6(1) and the corresponding obligations applied from 2 August 2027.

Article 50 sits in Chapter IV. It was not in those exceptions, so it took the general date of 2 August 2026. Article 50(6) already said that paragraphs 1 to 4 “shall not affect the requirements and obligations set out in Chapter III.”

Regulation (EU) 2026/1744 replaces Article 113, third paragraph, point (c). Chapter III, Sections 1, 2 and 3, with the exception of Article 6(5), now apply from:

Recital (40) of the Omnibus states that the initial application date for those Chapter III sections was 2 August 2026, and that keeping that date was no longer justified given delayed standards, common specifications, guidance, and national competent authorities.

Pure Global comparison (from the two Official Journal texts): Annex III high-risk Chapter III rules move from 2 August 2026 to 2 December 2027. Annex I / Article 6(1) rules, which already had a later clock, move from 2 August 2027 to 2 August 2028. Article 50’s 2 August 2026 application is not replaced.

Recital (17) names Regulations (EU) 2017/745 and 2017/746 as examples of the Union harmonisation legislation in Annex I, Section A. Fimea’s device notice says the 2 August 2028 Annex I date “also applies to medical devices when an AI system contained in them meets the conditions for being high risk laid down in Article 6.”

MDR and IVDR obligations are not postponed by this regulation. A CE-marked device still has to meet those regulations on their own terms.

Article 50 is in force now

Article 50 is titled “Transparency obligations for providers and deployers of certain AI systems.” Among other things it requires:

• providers to ensure that AI systems intended to interact directly with natural persons inform those persons that they are interacting with an AI system, unless that is obvious to a reasonably well-informed, observant and circumspect person in the context of use (Article 50(1), with a law-enforcement exception);
• providers of systems, including general-purpose AI systems, that generate synthetic audio, image, video or text to mark outputs as artificially generated or manipulated, in a machine-readable and detectable way, so far as technically feasible (Article 50(2), with listed exceptions);
• deployers of emotion-recognition or biometric-categorisation systems, and of systems that generate or manipulate deep fakes or certain public-interest text, to give the disclosures in Article 50(3) and (4).

Information under paragraphs 1 to 4 must be clear and distinguishable at the latest at the first interaction or exposure, and must meet applicable accessibility requirements (Article 50(5)).

Fimea states that these obligations also apply in the medical-device sector, that they do not depend on high-risk classification, and that they may already apply if a device AI system has one of the functionalities or uses in Article 50.

The Commission published guidelines on Article 50 on 20 July 2026. That page also states that the transparency obligations apply from 2 August 2026. This article does not restate the guideline PDF.

The Omnibus does add one transitional marking deadline. New Article 111(4) of the AI Act: providers of AI systems, including general-purpose AI systems, generating synthetic audio, image, video or text that have been placed on the market before 2 August 2026 shall take the necessary steps to comply with Article 50(2) by 2 December 2026. Recital (38) calls this a four-month transitional period so as not to disrupt the market. It is not a delay of Article 50(1), (3), (4) or (5), and it is not a delay of Article 50(2) for systems first placed on the market on or after 2 August 2026.

Who is “high-risk” under Article 6(1)

Article 6(1) is the Annex I route: an AI system is high-risk where it is a safety component of a product covered by Annex I legislation, or is itself such a product, and that product must undergo a third-party conformity assessment under that legislation.

The Omnibus amends the definition of “safety component” in Article 3, point (14), so that a component fulfils a safety function where its intended purpose is to prevent or mitigate risks to the health and safety of persons or property.

It also inserts Article 6(1a) to (1c). Article 6(1a) says that, for this Regulation including paragraph 1, AI systems that are solely used for non-safety related aspects of user assistance, performance optimisation, service efficiency, automation or convenience or quality control shall not qualify as safety components. Article 6(1b) adds that, notwithstanding 1a, AI systems whose failure or malfunctioning would endanger health and safety shall qualify as safety components.

Pure Global reading (limited to that wording): an algorithm sitting in a medical device is not high-risk under Article 6(1) merely because it is embedded. The intended purpose and the effect of failure still have to meet the amended tests. Whether a given product meets Article 6(1) remains a product-specific determination under the AI Act and the applicable MDR/IVDR conformity-assessment route. This article does not classify individual devices.

What manufacturers should do now

Derived from the dates and duties above, not from a new inspection campaign:

1. Treat Article 50 as current law for any device AI system that has an Article 50 functionality or use. High-risk delay is not a transparency delay. Use the Commission guidelines and Fimea’s device note as orientation; the binding text is Article 50.
2. Close the Article 50(2) marking gap by 2 December 2026 if a generative system was already on the market before 2 August 2026.
3. Map each AI function to Article 6(1)/Annex I versus Article 6(2)/Annex III under the amended safety-component definition. The two clocks are now 2 August 2028 and 2 December 2027.
4. Do not treat 2 August 2028 as extra time for MDR/IVDR clinical, QMS, or CE-marking work.
5. Correct any internal tracker that still uses 2 August 2027 as the Annex I high-risk date, or that still treats the Omnibus as a proposal.

The authentic English PDF is on the Publications Office Cellar: Regulation (EU) 2026/1744. ELI: http://data.europa.eu/eli/reg/2026/1744/oj. Fimea’s notice is here. For CE marking under the sectoral regulations, see Pure Global’s European Union market page.