---
title: "EN 18286: the AI Act's first harmonised standard"
url: https://zerofive.ai/en/blog/compliance/en-18286-harmonised-standard-ai-act
canonical: https://zerofive.ai/en/blog/compliance/en-18286-harmonised-standard-ai-act
language: en
published: 2026-10-16
updated: 2026-10-16
author: "ZeroFive.AI"
tags: EN 18286, AI Act, harmonised standards, quality management system, ISO IEC 42001
abstract: "EN 18286 has been in the EU Official Journal since 14 October 2026: Article 17 presumption, scope, link with ISO 42001 and questions for suppliers."
---

# EN 18286: the AI Act's first harmonised standard

**In brief.** On 14 October 2026 the European Commission published in the Official Journal of the EU Implementing Decision (EU) 2026/2274, which lists EN 18286:2026 as a harmonised standard under the AI Act. It is the first one. From that day, a provider that builds its quality management system according to this standard benefits from a presumption of conformity with the requirements of Article 17, within the limits the standard itself sets. The high-risk obligations still apply from December 2027 and August 2028, but there is now a recognised technical reference to build the work on.

Until this week Article 17 said what the quality management system of a high-risk AI provider must contain, thirteen items from the compliance strategy to resource management, without saying how to organise it. Every company could read it its own way, and every reading then had to be defended before a notified body or a market surveillance authority. EN 18286 closes that gap, at least for part of the regulation.

## The 14 October decision

[Implementing Decision (EU) 2026/2274](https://eur-lex.europa.eu/eli/dec_impl/2026/2274/oj) was adopted on 12 October, published in the L series of the Official Journal on the 14th and entered into force the same day. The text is short: two articles and an annex with a single reference, EN 18286:2026, "Artificial intelligence, Quality management system for EU AI Act regulatory purposes".

The mechanism is the one in Article 40 of the AI Act. High-risk systems that conform to harmonised standards whose references are published in the Official Journal are presumed to conform to the requirements those standards cover. Publication by CEN and CENELEC, at the end of July after ratification on 12 July, was not enough: the Commission had to verify the standard against its standardisation request, decision C(2025) 3871, and publish the reference. Until 13 October, applying EN 18286 meant following European good practice; since the 14th it means applying a harmonised standard.

The standard was drafted by the joint technical committee CEN/CLC/JTC 21 and answers point 9 of Annex I of the Commission's request, on quality management systems for providers, including post-market monitoring processes. The text is paid and is sold by the national standardisation bodies that are members of CEN.

## Content and structure of EN 18286

The standard addresses providers of AI systems of any size and sector, first of all those placing high-risk systems on the market or putting them into service. Deployers are not direct addressees: the standard reaches them through the obligations the provider owes them, such as instructions for use and channels to report incidents.

One definitional choice shapes everything else. In EN 18286 "quality" means compliance with the regulatory requirements of the AI Act, including the protection of health, safety and fundamental rights, while customer satisfaction, central to ISO 9001, falls outside the scope. In the same way, non-compliance means failing to meet an applicable regulatory requirement.

The structure follows the ISO management system layout, with a larger block on the life cycle of each individual system.

| Clause | Main content |
|---|---|
| 4. Quality management system | applicable regulatory requirements, scope, compliance strategy, documented information |
| 5. Leadership | top management commitment, quality policy, roles and responsibilities |
| 6. Planning | risks to the functioning of the management system, objectives |
| 7. Support | resources, competence, communication with authorities and notified bodies |
| 8. AI system realisation | life cycle, risk management, design, verification and validation, data, retirement, identification, continuous learning, technical documentation |
| 9. Operations and control | placing on the market, support services, supply chain, modifications, post-market monitoring, serious incidents, non-compliance |
| 10. Performance evaluation | management review, planning of changes |

Some elements of clauses 8 and 9 deserve attention because they touch points that projects often leave uncovered. Retirement is treated as a life cycle stage, with a duty to tell every deployer which measures to take when a system is withdrawn. Every version that goes through conformity assessment gets a unique reference, and every system placed on the market must remain identifiable along the distribution chain. Systems that keep learning after release must have their planned changes documented at design time, verified and validated one by one and reflected in the technical documentation.

On serious incidents the standard follows the deadlines of Article 73 of the regulation: two days for cases involving critical infrastructure, ten in the event of a death, fifteen in all other cases, counted from when the provider becomes aware of the event. On risk management, clause 8.2 requires a system covering the whole life cycle and refers, for the process itself, to the dedicated standard still in development, prEN 18228.

The annexes complete the picture. Annex A, informative, covers stakeholder engagement on risks to health, safety and fundamental rights. Annexes B and C contain the correspondence tables with ISO 9001:2015 and ISO/IEC 42001. Annex ZA links the clauses to the articles of the AI Act and is the one that sets the scope of the presumption.

## The scope of the presumption of conformity

The presumption applies to the requirements mapped in Annex ZA, and the scope is narrow by design. It covers Article 17(1), all points from a to m, and the first sentence of Article 11(1), the obligation to draw up technical documentation before placing on the market and keep it up to date.

| Covered by the presumption | Outside the presumption |
|---|---|
| Article 17(1), points a-m: the thirteen components of the management system | Article 17(2)-(4): proportionality, sectoral regimes, financial institutions |
| Article 11(1), first sentence: drawing up and updating technical documentation | Content of the Annex IV technical documentation |
| Existence of a risk management system as part of the quality system (point g) | The Article 9 risk management process |
| | Technical requirements on data, accuracy, robustness, cybersecurity, logging, human oversight |

The practical consequence is worth keeping in mind when press releases mention "presumption of conformity". A provider applying EN 18286 shows that it has a management system organised the way the AI Act requires. It does not thereby show that its system is accurate, robust or secure: those requirements wait for their own harmonised standards, and meanwhile must be demonstrated by other means, as point e of Article 17 itself provides.

The presumption is rebuttable and does not replace conformity assessment. A surveillance authority can challenge it with contrary evidence, and the management system has to work in practice. We saw this when writing about [Italian Decree 160/2026 on civil liability](/en/blog/compliance/ai-documentation-as-evidence-decree-160-2026): an approved procedure passes an audit, but in a specific case what counts is the record of what was done.

One design detail separates EN 18286 from ISO management system standards. The standard has no continual improvement obligation: it asks for the effectiveness of the system to be maintained, checked through management review. This fits its purpose, which is ensuring compliance with a regulation, and it changes how a company used to ISO 9001 sets objectives and indicators.

## Relationship with ISO 9001 and ISO/IEC 42001

For many companies the work starts from Annexes B and C. Organisations with an ISO 9001 certified quality system or an AI management system under [ISO/IEC 42001](/en/blog/compliance/iso-42001-ai-management-system) find the clause-by-clause correspondence and can focus on the differences. The tables map clauses, not concepts: they are a navigation aid, not a statement of equivalence.

The two ISO standards and EN 18286 serve different purposes. ISO/IEC 42001 is a certifiable international standard, centred on the organisation and applicable to any use of AI, including by companies that buy and deploy it. EN 18286 is a European standard centred on the product, written to implement Article 17 and to carry the presumption of conformity. The Commission had assessed that the objectives of ISO/IEC 42001 were not aligned with the management system required by the AI Act, which is why the European committee wrote a dedicated standard rather than harmonising the existing one.

Companies already certified to ISO/IEC 42001 therefore have a head start on the governance layer (policy, roles, competence, document control) and work to do on the product side: version identification, retirement, supply chain, serious incidents, non-compliance handling with withdrawal or recall. Those operating in sectors with a regulated quality system, such as medical devices under EN ISO 13485, can extend that system instead of building a second one in parallel, and the standard is written to allow it. Our [Accredia map](/en/blog/compliance/iso-42001-certification-italy-accredia) lists the bodies accredited in Italy for ISO/IEC 42001.

## Dates and the Italian framework

The [Digital Omnibus](/en/blog/compliance/digital-omnibus-regulation-2026-1744), Regulation (EU) 2026/1744, set the application dates of high-risk obligations at 2 December 2027 for Annex III systems and 2 August 2028 for those embedded in products regulated under Annex I. It also rewrote Article 17(2): implementation of the management system must be proportionate to the size of the provider, in particular for SMEs, start-ups and small mid-caps, while keeping the required degree of rigour and level of protection. SMEs without linked enterprises will be able to meet some elements in a simplified way, based on Commission guidelines that are still being prepared.

In Italy, [Legislative Decree No 179 of 7 October 2026](https://www.gazzettaufficiale.it/eli/id/2026/10/08/26G00200/SG), published in Official Gazette No 234 of 8 October and in force from 23 October, completes the national penalty framework. Breaches of provider obligations, including the Article 16 duty to have a quality management system complying with Article 17, fall in the band of up to EUR 15 million or 3% of turnover. The decree does not refer to the Digital Omnibus, and a penalty presupposes an applicable obligation: for high-risk systems that obligation starts on the European dates.

The other harmonised standards will follow, with CEN and CENELEC pointing to late 2026 and early 2027 for the texts, plus the Commission's assessment before publication in the Official Journal. They include risk management, conformity assessment, cybersecurity, data quality and logging. EN 18286 is the first of the series, and the others plug into it, because the management system is what calls them up and holds them together. The full calendar is in our article on [what happens on 2 December 2027](/en/blog/compliance/ai-act-december-2-2027-fourteen-months).

## Questions to ask suppliers

Most companies use AI systems built by others, and for them EN 18286 is not an obligation. It is still useful because it describes what should exist on the supplier side when buying a system that decides about customers, people or critical processes. Before signing or renewing a contract for a system that falls, or might fall, under [Article 6](/en/blog/compliance/high-risk-ai-systems-article-6-ai-act), it is worth asking:

- whether the supplier has a quality management system and against which reference, EN 18286, ISO/IEC 42001 or a sectoral system;
- how it identifies system versions and how it notifies the customer of changes, particularly for models that keep learning after release;
- which channel it provides for reporting incidents and malfunctions, and with what response times;
- which measures it foresees on retirement or withdrawal, and what the customer will have to do in that case;
- how it manages its own supply chain, starting with the third-party models the system is built on;
- which parts of the technical documentation and logs it makes available to the customer, including in case of litigation.

The answers serve two purposes. They show how ready the supplier is for 2027, and so how likely the system is to stay usable without interruption. They also show where the residual risk sits, which for a deployer translates into contract clauses and internal controls. A deployer becomes a provider if it substantially modifies a system, markets it under its own brand or uses it for a purpose other than the declared one, and in that case EN 18286 applies to it directly.

## Where to start

The prerequisite is the same as for any AI Act work: knowing which systems are in use and in which role. Without an [AI system inventory](/en/blog/compliance/ai-system-inventory-iso-42001) and a risk classification there is no way to establish whether EN 18286 is a standard to apply, a reference for suppliers, or both. For those who develop, the next step is a gap analysis between the existing management system and the clauses of the standard, starting from the correspondence tables in Annexes B and C.

In our [AI Rating](/en/services/ai-rating) the AI Risk dimension measures risk governance, compliance and traceability, and from today EN 18286 offers a recognised yardstick for it. To discuss your position you can [book an assessment call](https://calendly.com/fabiolalli/zerofive), [start the self-assessment](/en/start-ai-rating) or write to hello@zerofive.ai.

*This article describes the regulatory framework and its organisational implications, and is not legal advice. The text of EN 18286 is sold by national standardisation bodies and should be consulted in its official version.*

**Sources.** Commission [Implementing Decision (EU) 2026/2274](https://eur-lex.europa.eu/eli/dec_impl/2026/2274/oj) of 12 October 2026, OJ L 2026/2274 of 14.10.2026; [CEN-CENELEC, announcement of EN 18286:2026](https://www.cencenelec.eu/news-events/news/2026/en-in-the-spotlight/2026-07-30-ai-quality-management/), 30 July 2026; Regulation (EU) 2024/1689, Articles 11, 16, 17, 40, 73, as amended by [Regulation (EU) 2026/1744](https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=OJ:L_202601744); [Legislative Decree No 179 of 7 October 2026](https://www.gazzettaufficiale.it/eli/id/2026/10/08/26G00200/SG), Official Gazette No 234 of 8 October 2026; [European Commission, AI Act standardisation](https://digital-strategy.ec.europa.eu/en/policies/ai-act-standardisation).