EN 18286 clauses 9.6 and 9.7: serious incidents and non-compliance

EN 18286:2026 is a voluntary European Standard for the quality management system (QMS) of a provider of AI systems. Its clause 9.6 is about the report of serious incidents. Its clause 9.7 is about non-compliance of AI systems on the market or in service, and about AI systems presenting a risk. The requirements of EN 18286 apply only when a provider uses the standard.

For providers of high-risk AI systems, the legal duties are in the EU AI Act: Article 73 for serious incidents and Article 20 for corrective actions. Post-market monitoring in clause 9.5 is one source of information for the clause 9.7 procedures (see Lifecycle and operations). The dates from which the EU AI Act duties apply are on the EU AI Act pages. Modulos holds clauses 9.6 and 9.7 in the Organization Framework OFF-19 (see Operationalizing EN 18286 in Modulos).

Clause 9.6 — Reporting serious incidents

EN 18286 clause 9.6 tells the provider how to report serious incidents to the competent authorities. Clause 3.2.8 of EN 18286 uses the definition of a serious incident in Article 3(49) of the EU AI Act.

Investigation and report (9.6.1). EN 18286 requires the provider to have a process for the investigation of each serious incident. The process finds if there is a causal link between the AI system and the serious incident, or a “reasonable likelihood” of a causal link. EN 18286 requires a report to the competent authorities after the provider establishes a causal link. It also requires a report after the provider considers that a causal link is “reasonably plausible”.

Time scales (9.6.1). EN 18286 requires the provider to send the report immediately, and not later than these time scales:

Serious incident	Time scale in EN 18286 clause 9.6.1
A serious incident related to critical infrastructure	2 days
A serious incident with the death of a person	10 days
All other serious incidents	15 days

Each time scale starts when the provider becomes aware of the serious incident. The provider is permitted to send a first report that is provisional and not complete, followed by a complete report.

The legal time limits: EU AI Act Article 73. EN 18286 is voluntary. For “Providers of high-risk AI systems placed on the Union market”, Article 73 of the EU AI Act gives the legal time limits. Article 73 is different from clause 9.6.1 on these points:

• Authority that receives the report. Article 73(1) tells the provider to report to “the market surveillance authorities of the Member States where that incident occurred”. Clause 9.6.1 gives the competent authorities.
• The time limit of 2 days. Article 73(3) applies it to “a widespread infringement or a serious incident as defined in Article 3, point (49)(b)”. Point (49)(b) is “a serious and irreversible disruption of the management or operation of critical infrastructure”. Clause 9.6.1 applies its 2 days to each serious incident related to critical infrastructure. It does not name a widespread infringement. Article 73(3) requires the report “immediately, and not later than two days after the provider or, where applicable, the deployer becomes aware of that incident”.
• Death of a person. Article 73(4) requires the report immediately after the provider or the deployer establishes a causal relationship, or “as soon as it suspects” one. The time limit is not later than 10 days. Clause 9.6.1 gives the same 10 days, with the causal-link condition of the standard (see Causal link below).
• Causal link. Clause 9.6.1 uses the condition in Investigation and report above for all three time scales. Article 73(2) requires the report 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, not later than 15 days after the provider or, where applicable, the deployer, becomes aware of the serious incident”. Article 73(3) starts with “Notwithstanding paragraph 2” and has no causal-link condition. The Article 3(49) definition applies to point (b) too: a serious incident is “an incident or malfunctioning of an AI system that directly or indirectly leads to” the result in that point.
• Who becomes aware. In Article 73(2) to (4), each time limit starts when “the provider or, where applicable, the deployer” becomes aware of the serious incident. Clause 9.6.1 starts its time scales only from the awareness of the provider.
• Report that is not complete. Article 73(5) lets “the provider or, where applicable, the deployer” send “an initial report that is incomplete”, followed by a complete report, “Where necessary to ensure timely reporting”. Clause 9.6.1 gives its provisional report without this condition.

The full Article 73 duties, with the general time limit of 15 days, are in the Article 73 section of EU AI Act post-market monitoring.

Guidance from the Commission. A NOTE to clause 9.6.1 refers to Commission guidance and a reporting template for the report of serious incidents. On 26 September 2025, the Commission published “draft guidance and reporting template on serious AI incidents”, for the reports that Article 73 requires. The consultation for feedback on the two documents closed on 7 November 2025. The page of this consultation, changed last on 4 November 2025, gives the guidance as a draft.

Procedures (9.6.2). EN 18286 requires the provider to document, implement, and maintain procedures that report serious incidents in the time scales of 9.6.1 (9.6.2.1). These procedures include procedures with which deployers report serious incidents to the provider and suspend the use of the AI system. EN 18286 recommends that the procedures also include these items (9.6.2.2):

• The internal contacts that are responsible, and the internal escalation process.
• Awareness of serious-incident risks and of the escalation process for the applicable personnel.
• Processes that let the provider report in the time scales.
• Sufficient resources, for example competent personnel and the necessary equipment, for the investigation and to give information to the authorities.
• Documented information about each serious incident, its investigation, its results, and the steps that make the provider compliant again and keep it compliant.
• The procedures and obligations between the provider and the deployer that let the deployer report to the provider.
• The identification of each related AI system in the QMS scope that can cause the same serious incident.

Annex ZA (9.6). Table ZA.1 of EN 18286 shows clauses 9.6.1 and 9.6.2.1 against Article 17(1)(i) of the EU AI Act: “procedures related to the reporting of a serious incident in accordance with Article 73”. Table ZA.1 does not show clause 9.6.2.2, and it has no row for Article 73.

In Annex ZA, a presumption of conformity from these clauses applies from the citation of EN 18286 in the Official Journal. By Implementing Decision (EU) 2026/2274, the Commission published the reference on 14 October 2026 (see Harmonized standards). For each row and its conditions, see Annex ZA and presumption of conformity.

In Modulos. The Requirement for clause 9.6 is in the Organization Requirements table of Operationalizing EN 18286 in Modulos. One serious incident can also start the clause 9.7 steps. Clause 9.6.2.2 a) of EN 18286 recommends an internal escalation process for reports of serious incidents. Modulos recommends that the provider connect this process to its procedures for clause 9.7.

Clause 9.7 — Non-compliance

EN 18286 clause 9.7 tells the provider how to find and correct non-compliance of AI systems on the market or in service. It also gives the steps for AI systems presenting a risk.

Non-compliance (3.1.15). In EN 18286 clause 3.1.15, non-compliance is non-fulfilment of applicable regulatory requirements. Clause 9.7 uses non-compliance for AI systems on the market or in service.

The legal duty: EU AI Act Article 20(1). For high-risk AI systems, the legal duty to do corrective actions comes from the EU AI Act. Article 20(1) tells providers of high-risk AI systems this:

Providers of high-risk AI systems which consider or have reason to consider that a high-risk AI system that they have placed on the market or put into service is not in conformity with this Regulation shall immediately take the necessary corrective actions to bring that system into conformity, to withdraw it, to disable it, or to recall it, as appropriate. They shall inform the distributors of the high-risk AI system concerned and, where applicable, the deployers, the authorised representative and importers accordingly.

Clause 9.7.1 of the voluntary standard adds a documented process. Clause 7.3.2.2 of EN 18286 requires the provider to inform the related interested parties of non-compliance, of AI systems presenting a risk, and of serious incidents, as applicable regulatory requirements tell it to. The NOTEs to clause 7.3.2.2 refer to Articles 20(1), 20(2), and 73 (see The QMS clauses).

Procedures and steps (9.7.1). EN 18286 requires the provider to document, implement, and maintain procedures that find if AI systems on the market or in service do not comply with the applicable regulatory requirements. If a non-compliance occurs, EN 18286 requires the provider to do these steps:

• a) The provider finds how large the non-compliance is, and its severity. In an EXAMPLE, the provider finds if the non-compliance has an effect on other AI systems in the same QMS.
• b) The provider examines if it is necessary to remove the causes. This is to make sure that the non-compliance does not occur again, in the same area or in a different area. For this, the provider examines the non-compliance and finds its causes. The provider also finds if other non-compliances of the same type occur or can occur.
• c) The provider immediately does the applicable corrective actions: bring into compliance, withdraw, disable, or recall the AI system.
• d) The provider examines the effectiveness of each corrective action that it does.
• e) The provider examines the data about non-compliance to find trends, systemic problems, or new risks. From this data, the provider finds if action is necessary to prevent problems.
• f) If necessary, the provider changes the QMS (see clause 10), the risk management system, the design of the AI system, the deployment conditions, or the controls of operation.

EN 18286 also requires the provider to inform the deployers, distributors, authorized representatives, and importers, as applicable.

Selection of a corrective action. EN 18286 gives no criteria to select which of the corrective actions in clause 9.7.1 c) apply. A provider can write criteria in its procedures, for example who makes the decision. These criteria are a decision of the provider, not a requirement of EN 18286. Clause 9.7.1 c) requires the provider to do the applicable corrective actions immediately.

AI systems presenting a risk (9.7.2). The EU AI Act has this text in Article 79(1):

AI systems presenting a risk shall be understood as a ‘product presenting a risk’ as defined in Article 3, point 19 of Regulation (EU) 2019/1020, in so far as they present risks to the health or safety, or to fundamental rights, of persons.

EN 18286 clause 3.2.9 uses the Regulation (EU) 2019/1020 definition, and its Note 1 gives the same limit as Article 79(1). In that definition, the possible bad effect of the AI system goes above a level that is “reasonable and acceptable”. The test looks at the intended purpose, or at conditions of use that are “normal or reasonably foreseeable”. It includes how long the AI system is in use. Where applicable, it also includes the requirements for how the AI system is put into service, installed, and maintained.

EN 18286 requires the provider to have these procedures and steps for AI systems presenting a risk:

• Procedures (9.7.2.1). Documented procedures find if an AI system on the market or in service is an AI system presenting a risk.
• Investigation (9.7.2.1). When the provider becomes aware of such an AI system, it makes an investigation of the causes immediately. Where applicable, it does this together with the deployer that sent the report.
• Information to authorities (9.7.2.2). The provider informs the market surveillance authorities that are competent for the AI system. Where applicable, it also informs the notified body that did the conformity assessment of the AI system. As a minimum, the information gives the type of risk and the related corrective actions of the provider.

For high-risk AI systems, the legal duty is Article 20(2) of the EU AI Act:

Where the high-risk AI system presents a risk within the meaning of Article 79(1) and the provider becomes aware of that risk, it shall immediately investigate the causes, in collaboration with the reporting deployer, where applicable, and inform the market surveillance authorities competent for the high-risk AI system concerned and, where applicable, the notified body that issued a certificate for that high-risk AI system in accordance with Article 44, in particular, of the nature of the non-compliance and of any relevant corrective action taken.

Thus Article 20(2) is different from clause 9.7.2.2 on two points. It gives “the notified body that issued a certificate for that high-risk AI system in accordance with Article 44”. Its information includes “the nature of the non-compliance”, and clause 9.7.2.2 gives the type of risk.

Documented evidence (9.7.3). EN 18286 requires documented information that shows the type of each non-compliance and the steps after it. The documented information also shows the results of each corrective action.

Annex ZA (9.7). Table ZA.1 of EN 18286 shows clause 9.7 and its subclauses against three rows of Article 17(1) of the EU AI Act:

Article 17(1) row in Table ZA.1	EN 18286 clauses for clause 9.7
First sentence: providers “shall put a quality management system in place that ensures compliance with this Regulation”	9.7
Point (c): “the development, quality control and quality assurance of the high-risk AI system”	9.7.1, 9.7.2.1, and 9.7.3
Point (j): “the handling of communication with national competent authorities, other relevant authorities, including those providing or supporting the access to data, notified bodies, other operators, customers or other interested parties”	9.7.2.2, with clause 7.3

The remark for the first sentence gives this limit: the use of EN 18286 is not sufficient to make sure that the provider complies with all regulatory requirements of the EU AI Act. Table ZA.1 shows more clauses for the first sentence and for point (c) than the table above. Annex ZA shows that compliance with the clauses of these rows gives a presumption of conformity, in the scope of the standard, from the citation of 14 October 2026. Each row, with the full remark, is on Annex ZA and presumption of conformity.

In Modulos. For the Requirement that holds clause 9.7, see the OFF-19 table on Operationalizing EN 18286 in Modulos.

How clauses 9.6 and 9.7 work together

EN 18286 clauses 9.6 and 9.7 can apply at the same time. A serious incident can also show that an AI system does not comply with an applicable regulatory requirement. Then it can be necessary for the provider to send the 9.6 report and do the 9.7 steps at the same time.

The investigation of a serious incident can also have an effect on the QMS. Clause 10.1.1.1 of EN 18286 requires one more management review when the investigation of a serious incident shows that the QMS, or the measures of the QMS, are not sufficient (see The QMS clauses).

In EN 18286, deployers give information to the provider and receive information from the provider:

• From the deployer. Clause 9.6.2.1 requires procedures with which deployers report serious incidents and suspend use. Clause 9.2 recommends a mechanism with which each deployer can send the feedback that it receives about possible risks for health, safety, and fundamental rights.
• To the deployer. Clause 9.7.1 requires information to deployers, as applicable. Clause 8.6 requires information to the deployer about all the retirement measures that are necessary for the deployer to do. Clause 9.5.5 requires monitoring requirements in the instructions for use when the provider cannot find a non-compliance without the deployer, if this is necessary to decrease risk.

Related pages

EN 18286 overview

Structure, Annex ZA, citation, and the Modulos Frameworks of EN 18286

The QMS clauses

Communication (7.3) and management review (10.1)

Lifecycle and operations

Post-market monitoring (9.5)

EU AI Act post-market monitoring

The Article 72 and Article 73 obligations of the Regulation

Sources of this page

• EN 18286:2026 (European Standard of July 2026): clauses 3.1.15, 3.2.8, 3.2.9, 7.3.2.2, 8.6, 9.2, 9.5.5, 9.6, 9.7, and 10.1.1.1, and Annex ZA (Table ZA.1).
• Regulation (EU) 2024/1689 (EU AI Act), consolidated text of 27 July 2026: Articles 3(49), 17(1), 20, 73, and 79(1). EUR-Lex.
• The Commission Implementing Decision (EU) 2026/2274 of 12 October 2026 (OJ L, 2026/2274): Article 1 and the Annex, with the reference of EN 18286:2026. EUR-Lex.
• The consultation page of the European Commission, “AI Act: Commission issues draft guidance and reporting template on serious AI incidents, and seeks stakeholders' feedback”, published 26 September 2025, changed last on 4 November 2025. Digital Strategy.

Disclaimer

This page gives general information only. It is not legal advice.