Emergency triage

Core judgement

The AI Act does not prohibit emergency triage AI. Its high-risk list covers systems that evaluate or classify emergency calls, dispatch or prioritise police, firefighters or medical aid, or triage patients in emergency healthcare.¹ Following an enacted amendment, the main duties begin on 2 December 2027 for systems classified as high-risk. Under Article 6(3), qualifying narrow or preparatory systems posing no significant risk and no material influence on outcomes may be treated as non-high-risk, while profiling systems always remain high-risk.¹²

Article 86 has applied since 2 August 2026, but whether its Annex III trigger operates before Article 6 classification rules apply on 2 December 2027 remains unresolved.¹ If triggered, and only where EU law provides no equivalent explanation right, the emergency-service or healthcare deployer must explain AI’s role and the main elements of a decision legally or similarly significantly and adversely affecting a person’s health, safety or fundamental rights. The market-surveillance authority enforces it and can receive Article 85 complaints.¹ The decision need not be solely automated.

Since 25 May 2018, Article 22 has restricted significant decisions based solely on automated processing of personal data.³ The emergency-service or healthcare controller bears the duty; data-protection authorities and courts enforce it. A genuine decision by a dispatcher or clinician may place the outcome outside Article 22, depending on the facts.⁴ If a solely automated decision uses health data, Articles 9 and 22 require the controller to meet special-category conditions and provide suitable safeguards, enforced by data-protection authorities and courts since that date.⁵

From 2 December 2027, Article 26 requires the emergency-service or healthcare deployer to tell a person when high-risk AI makes or assists a decision about them. It also requires competent, trained and authorised human oversight. Articles 14 and 16 require the provider to design the system so overseers can understand its limits, monitor it, disregard or reverse outputs, intervene and stop it safely. The designated market-surveillance authority, or the AI Office where Article 75 assigns it competence, enforces the provider duty; the market-surveillance authority enforces deployer duties.¹

Article 27 will require a public authority or private entity providing public services to assess fundamental-rights effects before first deploying a covered emergency system and notify the market-surveillance authority. That authority enforces the duty from 2 December 2027.¹

Boundary. These regulatory duties do not give a caller or patient an immediate right to a second triage assessment, rerouting or changed priority. Article 86 supplies no standard request form, response deadline, emergency procedure or appeal route. No official guidance supplied here explains how an AI-use notice should be delivered during an emergency. The applicable date also remains unresolved where one system is both emergency-triage AI and regulated medical-device AI.