The EU AI Act and emergency response systems: what changes, and when

What the EU AI Act regulates in outline

The EU AI Act is European regulation classifying AI systems on the basis of the risk they carry. The greater the potential impact of an AI application on people's safety, health or fundamental rights, the stricter the requirements placed on it. At one end of that spectrum sit applications with negligible risk, for which barely any additional rules apply. At the other end sit systems designated as high risk: those carry considerably heavier obligations, aimed at matters such as human oversight, the quality of underlying data and the traceability of decisions.

For organisations in emergency response and crisis management, the second category is especially relevant. The legislation names a number of application areas in which AI is deemed to present elevated risk, simply because the consequences of a wrong assessment there work through directly into people's safety.

Why emergency response and dispatch systems count as high risk

Under the EU AI Act, AI systems used by emergency call centres, and systems supporting dispatch decisions (which unit is sent where), fall into the high-risk category under Annex III. The same applies to AI used in triage: assessing and prioritising incoming calls by urgency, including emergency healthcare patient triage.

The logic behind that is easy to follow. These are systems where wrong output can have direct consequences for life and health. When an algorithm helps determine which call takes priority, which unit is dispatched, or how urgent a situation is judged to be, a direct link arises between the quality of that system and the safety of the public and of responders. Hence the legislator sets heavier requirements here than for AI supporting administrative tasks.

The shifted timeline: from August 2026 to December 2027

For these standalone high-risk systems, the original date on which the associated obligations would take effect was 2 August 2026. Through the so-called Digital Omnibus, that date has moved to 2 December 2027. What began as a political agreement between the Council and the European Parliament is now settled legislation: Regulation (EU) 2026/1744 was published on 24 July 2026 and entered into force on 27 July 2026. Organisations using or developing emergency call centre, dispatch or triage systems with AI components therefore have more time to meet the full set of obligations than originally foreseen. Note, however, that certain obligations, such as the prohibited AI practices and part of the transparency requirements, remain in force on their original dates.

Important to understand: this does not mean nothing applies until that date. Certain transparency obligations take effect earlier. The shifted timeline mainly creates room for the heavier, more intrusive obligations, not for the regulation as a whole. That the timeline was adjusted also shows this regulation is still very much in development: the precise elaboration and interpretation of what is expected of organisations will crystallise further in the period ahead.

For emergency response organisations this mainly means there is time to determine carefully which AI systems inside their own organisation fall into which category, and which steps that requires, rather than having to put something in place at short notice.

What this means for AI in crisis communication specifically

A question that logically arises: does AI used in crisis communication — drafting texts, summarising a situational picture, or translating a message — also fall under that same high-risk category? The answer is more nuanced than it first appears.

There is a material difference between AI that genuinely decides or co-decides (for instance which unit is dispatched, or how a call is prioritised) and AI that supports communication without taking that decision: preparing a draft text a person reviews and approves, summarising incoming signals, or translating a message into another language. The first type goes directly to the heart of why dispatch and triage systems count as high risk. For the second type that is less self-evident, because the eventual decision and accountability stay with a person.

Even so, this is not a distinction organisations may simply rely on. Whether an AI application is in practice purely supportive, or in fact becomes (co-)determining for decisions, depends on how the system is set up and used. Organisations are therefore well advised to be able to substantiate which category a specific application falls into, rather than assuming it. Precisely because the regulation is still moving, as the shifted timeline shows, care and a solid justification are wiser than a hasty conclusion in either direction.

Sources

• EU AI Act, Annex III: high-risk AI systems referred to in Article 6(2) — point 5(d) on emergency call classification, dispatch prioritisation and emergency healthcare triage.
• Regulation (EU) 2026/1744 (Digital Omnibus) — published 24 July 2026, in force 27 July 2026; moves the Annex III high-risk obligations to 2 December 2027.
• EENA, How the AI Act Will Shape the Future of Emergency Services.
• Rayvn, EU AI Act Emergency Dispatch: High Risk.