The Digital Omnibus on AI is now law. Regulation (EU) 2026/1744 of 8 July 2026 was published in the Official Journal on 24 July 2026 and entered into force on 27 July 2026. A week later, on 2 August 2026, the AI Act’s enforcement powers and its Article 50 transparency obligations started to apply, as the European Commission confirmed in press release IP/26/1714 on 31 July 2026.
In March we looked at what the proposed delay would mean for medical device manufacturers. The dates we discussed then are now confirmed. For an AI-enabled medical device or IVD, the high-risk requirements apply from 2 August 2028. If your software includes a chatbot or generates content, Article 50 applies already.
What the AI Omnibus fixed
The Omnibus amends Article 113 of the AI Act to set two new application dates for the high-risk rules in Chapter III:
- 2 December 2027 for AI systems classified as high-risk under Article 6(2) and Annex III (the standalone use cases, such as employment or access to essential services);
- 2 August 2028 for AI systems classified as high-risk under Article 6(1), which covers AI systems that are products, or safety components of products, regulated by the legislation listed in Annex I.
The MDR and the IVDR remain in Annex I, Section A. An AI-enabled device that requires notified body conformity assessment under the MDR or IVDR is therefore high-risk under Article 6(1), and 2 August 2028 is the date to plan against. The Omnibus moved the Machinery Regulation to Section B, but it did not move the MDR or IVDR. That move is part of the Commission’s December 2025 proposal to revise the MDR and IVDR, which we covered in Narrowing of the EU AI Act for Medical Devices and IVDs, and it remains a proposal.
The Omnibus also rewrites the AI literacy obligation in Article 4. Providers and deployers must now “take measures to support the development of AI literacy of their staff”, which is a lighter formulation than the original text.
What started on 2 August 2026
From 2 August 2026, the AI Office and national authorities can enforce the AI Act. On the same date, Article 50 began to apply. In the Commission’s words, chatbots and other interactive AI systems “will have to tell users they are dealing with AI, not a human”, deepfakes must be labelled, and AI-generated or altered content must carry machine-readable marks.
Two documents support these obligations. The final Code of Practice on transparency of AI-generated content was published on 10 June 2026. It is voluntary, and the Commission and the AI Board have confirmed it is an adequate tool to demonstrate compliance with Article 50(2), (4) and (5). On 20 July 2026 the Commission then approved its guidelines on the Article 50 transparency obligations (C(2026) 5054).
There is one transitional arrangement. Providers of generative AI systems placed on the market or put into service before 2 August 2026 have until 2 December 2026 to meet the marking and detection obligation in Article 50(2). The guidelines are clear that this is limited: a system that is both interactive and generative benefits only for the marking obligation, and the Article 50(1) disclosure obligation “must be ensured as of 2 August 2026.”
What Article 50 means for software as a medical device
The guidelines address several points that are directly relevant to medical device software.
Chatbots and conversational features (Article 50(1)). The guidelines list chatbots in healthcare among the systems in scope. A patient-facing symptom checker, a conversational companion app or a support assistant built into a device must tell users they are interacting with AI, at the latest at the time of first interaction. The guidelines add that periodic reminders are likely to be needed in riskier contexts, naming health advice and interactions with vulnerable persons.
The “obvious” exception. No disclosure is required where it is obvious to a reasonably well-informed, observant and circumspect person that they are dealing with AI. The guidelines say this exception should be interpreted restrictively. They give as an example of an obvious case “Interactive AI systems intended only to be used by properly trained health professionals to support medical diagnosis and suggest related treatments.” A clinician-only tool may therefore qualify, but the provider has to assess and document that conclusion. Where the general public or vulnerable users can reach the system, the exception cannot be relied on.
Generated content (Article 50(2)). Systems that generate synthetic text, audio, image or video must mark outputs in a machine-readable format. The guidelines list “AI-generated summaries of text” among the changes that require marking, which is relevant to clinical documentation and ambient scribe tools. Transcriptions of conversations fall under the exception, as does “Medical image technical processing or reconstruction within medical devices (e.g. annotations, boxes, contours, circles, heatmaps).”
Article 50 sits alongside the high-risk rules. The guidelines confirm that a system covered by Article 50 may also be high-risk and subject to those requirements. For an AI-enabled device, 2028 is the date for Chapter III, while any interactive or generative feature has to meet Article 50 today.
What to do next
- Map your AI features against Article 50. List every conversational interface and every feature that generates text, audio, images or video, whether patient-facing or clinician-facing. Record who can access each one.
- Confirm your role. If you place software on the market under your own name, you are the provider for Article 50, including where it runs on a third-party model accessed through an API.
- Check disclosure now. For interactive features, confirm users are told they are interacting with AI at first interaction. If you rely on the “obvious” exception for a clinician-only tool, write down the reasoning and the controls that keep the tool away from patients.
- Plan marking by 2 December 2026. For generative features placed on the market before 2 August 2026, agree a marking approach and consider whether to sign or align with the Code of Practice. Anything placed on the market from 2 August 2026 needed marking from day one.
- Keep the high-risk programme moving. Use the time to 2 August 2028 to build AI Act requirements into your MDR or IVDR quality system and technical documentation, and discuss AI competence and expectations with your notified body early.
Specculo supports software manufacturers with AI Act and MDR/IVDR obligations, from Article 50 scoping to building AI requirements into a compliant software lifecycle. Our SaMD developer guide covers the device side, and if you would like to talk through how the AI Act applies to your product, get in touch.
Sources
- Regulation (EU) 2026/1744 (Digital Omnibus on AI), Official Journal, 24 July 2026
- European Commission press release IP/26/1714: Commission starts enforcing AI Act rules and new transparency requirements on 2 August (31 July 2026)
- European Commission news item on the start of enforcement
- Guidelines on the transparency obligations for certain AI systems under Article 50 AI Act, C(2026) 5054 (20 July 2026)
- Code of Practice on transparency of AI-generated content
- Regulation (EU) 2024/1689 (AI Act)