Healthcare Triage AI Under the EU AI Act: Which Route Applies?
AI triage that is a medical device follows the Annex I / MDR route (2 Aug 2028). Public emergency triage of essential care is Annex III pt 5 (2 Dec 2027).
Almost all clinical triage AI is high-risk — but it gets there by two different doors, and they carry different deadlines, different conformity routes, and different lead regulators. If your triage tool is a medical device, it is high-risk through the Annex I product route (Article 6(1)), assessed under the Medical Devices Regulation (MDR), with the 2 August 2028 deadline. If a public authority or emergency service uses AI to dispatch or triage access to essential healthcare, that is high-risk through Annex III point 5 (Article 6(2)), with the 2 December 2027 deadline. Both EU AI Act, Regulation (EU) 2024/1689 — but you must know which door you are walking through.
If you build clinical triage software, or you are a hospital, ambulance service or public health body deploying it, this page tells the two routes apart and sets out what each owes.
Route one: triage AI that is a medical device — the Annex I product route
Most software that triages, prioritises or risk-stratifies patients on the basis of their clinical condition is a medical device under the MDR (Regulation (EU) 2017/745) — its intended purpose is diagnosis, prognosis, prediction or treatment-prioritisation. Article 6(1) of the AI Act makes an AI system high-risk when it is (or is a safety component of) a product covered by the EU harmonisation legislation listed in Annex I — which includes the MDR — and that product must undergo third-party conformity assessment.
What this route means in practice:
- The AI Act's high-risk requirements (Articles 9–15) are met through the MDR conformity assessment, integrated with the device's existing certification rather than run as a separate parallel track — generally via the MDR notified body.
- The deadline is 2 August 2028, not 2 December 2027. Product-embedded high-risk AI under Article 6(1) is on the later track.
- The lead regulatory frame is medical-device law (MDR, the notified body, clinical evaluation), with the AI Act's requirements layered in.
A symptom-checker that outputs a clinical urgency level, an ED triage model that risk-stratifies arrivals by acuity, or imaging software that prioritises a worklist by suspected pathology are typically MDR medical devices and travel this route. (The IVDR is the parallel regime where the triage rests on in-vitro diagnostic data; same logic, same 2 August 2028 deadline.)
Route two: public-authority dispatch and triage of essential care — Annex III point 5
The second door has nothing to do with whether the software is a medical device. Annex III point 5 covers access to and enjoyment of essential public services and benefits:
- point 5(a) — AI used by, or on behalf of, public authorities to evaluate eligibility for essential public services and benefits, including healthcare services, and to grant, reduce, revoke or reclaim them;
- point 5(d) — AI intended to be used for dispatching, or establishing priority in the dispatching of, emergency first-response services (police, firefighters, medical aid) and emergency healthcare patient triage systems.
So when a public ambulance service or emergency call centre uses AI to prioritise dispatch or triage who gets seen first, or a public health authority uses AI to decide eligibility for a publicly funded healthcare service, that is high-risk under point 5 — by virtue of the public-service-access function, regardless of medical-device status. This route runs on the stand-alone Annex III timetable: 2 December 2027 (set by the Digital Omnibus, Regulation (EU) 2026/1744).
| Triage scenario | Route | Deadline | Key frame |
|---|---|---|---|
| Symptom-checker / ED acuity tool that is an MDR medical device | Annex I product route (Art 6(1)) | 2 Aug 2028 | MDR + notified body |
| Imaging worklist-prioritisation device (MDR) | Annex I product route (Art 6(1)) | 2 Aug 2028 | MDR + notified body |
| Public ambulance AI prioritising emergency dispatch | Annex III point 5(d) | 2 Dec 2027 | Public-service access + FRIA |
| Public emergency-care patient-triage system | Annex III point 5(d) | 2 Dec 2027 | Public-service access + FRIA |
| Public health body AI deciding eligibility for funded care | Annex III point 5(a) | 2 Dec 2027 | Public-service access + FRIA |
A single deployment can touch both doors — an MDR-certified triage device used by a public emergency service to dispatch meets the device route on the product side and the point 5 obligations on the deployer side. Map each layer rather than forcing the whole system through one route.
What each route owes — and the FRIA that point 5 adds
Both routes require the substance of the high-risk regime: risk management (Article 9), data governance (Article 10), technical documentation (Article 11 / Annex IV), logging (Article 12), human oversight (Article 14), and accuracy, robustness and cybersecurity (Article 15) — a triage score must inform a clinician or dispatcher, never replace the human decision. The deployer (hospital, ambulance service, health authority) carries the Article 26 duties: operate per instructions, oversee, monitor, retain logs.
The decisive add-on on the point 5 route: because the deployer is a public authority (or a body acting on its behalf), an Article 27 FRIA is mandatory before first use — distinct from a GDPR Article 35 DPIA, though it can build on one under Article 27(4). The device route does not, by itself, trigger a FRIA; the public-service-access route does.
Deadlines, kept distinct. The Annex I device route (Article 6(1)) was originally timetabled at 2 August 2027 and now applies from a fixed 2 August 2028. The stand-alone Annex III route (Article 6(2)) was originally timetabled at 2 August 2026 and now applies from a fixed 2 December 2027. Both dates are law under the Digital Omnibus — Regulation (EU) 2026/1744, published in the Official Journal on 24 July 2026 and in force since 27 July 2026. Plan against them.
Worked example: a regional ambulance trust and its triage vendor
Rettungsdienst Mittelhessen, a public regional ambulance service, deploys an AI dispatch-priority tool in its emergency call centre to rank incoming calls by urgency. As a public emergency service triaging access to essential care, this is high-risk under Annex III point 5(d) — route two. The trust, a public-body deployer, must complete an Article 27 FRIA before go-live, assign Article 14 human oversight so dispatchers can override the ranking, and register under Article 49. Its target date is 2 December 2027 (originally 2 August 2026, moved by the Digital Omnibus).
The tool itself is supplied by Triagon GmbH, whose symptom-acuity engine is also CE-marked as an MDR medical device. For Triagon as provider of a medical-device AI, the high-risk requirements are met through the MDR conformity route — route one — on the 2 August 2028 track. So the same overall system sits on two timetables: Triagon's device certification under MDR by 2 August 2028, and the trust's public-service point 5 obligations by 2 December 2027. Each party scopes its own route; neither can assume the other's covers it. (The trust, as a public authority, has no SME cap; Triagon, if an SME, benefits from the Article 99(6) cap.)
How Confir helps
Confir registers each healthcare-triage AI use and routes it correctly: it tests whether the system is an Annex I product (MDR/IVDR device route, 2 August 2028) or a stand-alone Annex III point 5 public-service use (2 December 2027), derives the role (deployer under Article 26, provider under Article 16), and — for the public-service route — flags the Article 27 FRIA. The Article 9–15 requirements are surfaced for both routes, mapped to whichever conformity track applies. The engine is deterministic and rule-based, each finding cited to the Article or Annex point that fired.
Related guides
- Healthcare AI and the medical-device routes
- Emergency-services dispatch AI under Annex III point 5(d)
- Essential public-services AI under Annex III point 5
- Medical-imaging AI and the EU AI Act
- How Article 6 classifies high-risk AI
- The Fundamental Rights Impact Assessment (Article 27)
Manage your EU AI Act compliance in one place
Confir automates risk classification, technical documentation, and audit trails for any company. No consultants. No 6-month projects. 14-day free trial.
Start free trial →