Skip to main content

Explainer

EU AI Act Article 73 Serious Incident Reporting Requirements: The Clock Starts When You Find Out

By AInews Desk
EU AI Act Article 73 Serious Incident Reporting Requirements: The Clock Starts When You Find Out
6 min read
1168 words

Under the EU AI Act Article 73 serious incident reporting requirements, the deadline does not wait for your investigation to finish. It starts when the provider or deployer becomes aware of the serious incident. Article 73 does give fixed limits of 2, 10 and 15 days, but its main instruction is to report "immediately." The day counts set the latest possible deadline. By the end of this explainer, you will know what counts as a serious incident, who has to report it and to whom, how each deadline works, and what you can and cannot do to the system once a report is filed.

What counts as a "serious incident"?

Article 3(49) defines a serious incident as an incident or malfunctioning of an AI system that directly or indirectly leads to any of the following outcomes:

  • (a) the death of a person, or serious harm to a person's health
  • (b) a serious and irreversible disruption of the management or operation of critical infrastructure
  • (c) the infringement of obligations under Union law intended to protect fundamental rights
  • (d) serious harm to property or the environment

An AI system does not have to cause the harm on its own. It is enough for the system to be part of the chain of events that led to it. A credit-scoring model can meet the definition even though it never physically touches anyone. If it plays a part in breaching anti-discrimination obligations, it can fall under point (c).

The four categories also feed directly into the timelines. Point (b) triggers its own faster deadline, and point (c) is the one category that continues to apply under the sector-specific exemptions covered below.

Who reports, and to whom

Article 73(1) says providers of high-risk AI systems placed on the Union market shall report any serious incident to the market surveillance authorities of the Member States where that incident occurred. The Service Desk summarizes the duty this way: providers of high-risk AI systems must report incidents to national market surveillance authorities or to the AI Office, in cases where the enforcement falls within its scope under Article 75(1). The report does not stop at the national level. National competent authorities must immediately notify the Commission of any serious incident, whether or not they have taken action on it.

Deployers also have a role. Article 73 refers to the moment the provider "or, where applicable, the deployer" becomes aware of an incident. Providers and deployers should therefore agree on how incident information will pass between them. If a hospital or bank using the system learns of a problem first, the deadline may already be running before the provider hears about it.

The 2, 10 and 15-day deadlines, compared

Article 73(2) sets the 15-day rule. A report must be made immediately after the provider has established a causal link between the AI system and the serious incident or the reasonable likelihood of such a link, and in any event no later than 15 days after awareness. The same paragraph adds that the reporting period shall take account of the severity of the serious incident.

Two situations shorten that limit:

Incident typeTrigger for "immediately"Outer limit
Standard serious incidentCausal link or its reasonable likelihood15 days after awareness
Death of a personCausal link established or suspected10 days after awareness
Widespread infringement or 3(49)(b)Report immediately2 days after awareness

The bar for a death is lower than the general standard of "reasonable likelihood." The report is due as soon as it suspects a causal relationship. The two-day limit covers critical-infrastructure disruptions and "widespread infringements." Article 3 defines a widespread infringement partly as harm, or likely harm, to the collective interests of individuals residing in at least two Member States other than the Member State in which the act or omission originated.

Short deadlines and incomplete facts often go together, and the regulation allows for that. To keep reporting on time, a provider or deployer may submit an initial report that is incomplete, followed by a complete report. Waiting for a complete picture risks missing the deadline.

After the report: don't change the evidence

Reporting is the first step. Article 73(6) says that, after the report, the provider shall, without delay, perform the necessary investigations in relation to the serious incident and the AI system concerned. That work shall include a risk assessment of the incident, and corrective action.

One rule stands out. The provider must cooperate with the competent authorities and, where relevant, with the notified body. It also shall not perform any investigation which involves altering the AI system concerned in a way which may affect any subsequent evaluation of the causes of the incident, prior to informing the competent authorities of such action.

Air accident investigation is a useful comparison. Wreckage stays where it is until investigators have examined it, because moving it destroys evidence. Engineers usually want to patch a fault quickly, but this rule requires them to tell the authorities first. The rule does not ban changes. It sets the order: notify, then alter.

A quiet data center aisle in which one server cabinet stands closed off behind a plain red rope barrier while the surrounding racks run normally.A quiet data center aisle in which one server cabinet stands closed off behind a plain red rope barrier while the surrounding racks run normally.

Exemptions for systems already covered by other EU rules

Article 73 adjusts reporting obligations for AI systems covered by other Union legislation, as noted in the summary of reporting requirements. For high-risk AI systems listed in Annex III whose providers are subject to Union legislative instruments laying down reporting obligations equivalent to those set out in this Regulation, only point (c) incidents must be reported under the AI Act.

Medical technology gets similar treatment. Where a high-risk AI system is a safety component of a device, or is itself a device, under Regulations (EU) 2017/745 and (EU) 2017/746, notification is again limited to point (c) incidents. These reports go to the national competent authority chosen for that purpose by the Member States where the incident occurred.

Point (c) is the category that survives both exemptions, which suggests a deliberate division of work. Existing sectoral regimes already handle injuries, equipment failures and property damage. Fundamental-rights harms are the area the AI Act adds, so a medical-device maker that already reports patient harm elsewhere still needs a process for spotting discrimination or other rights breaches caused by its AI.

What to set up before an incident happens

The deadlines depend on awareness. The incomplete-report option lets a provider or deployer submit an initial report that is incomplete, followed by a complete report. The rule against altering the system means change-freeze procedures have to be ready before an incident, because there will be no time to write them during one.

Sources

6 min read
•
1168 words

Tags

EU AI Act Article 73 serious incident reporting requirementsAI serious incident notification timelineEU AI Act 2 10 15 day reporting deadlinesAI incident reporting obligations providers

Related Articles