AI Standardisation Watch
Where the European standards supporting the EU AI Act stand today, and what that means legally. Organised by the questions readers ask. Updated after each newsletter.
Status as of 15 September 2026 · Based on Inclusiveness Newsletter, Editions 1–16.
Each standard has its own page with status, presumption gate, voting history, intended audience, sources and machine-readable data. Published standards first.
A working group writes a first draft against the Commission's request.
National members comment. Thousands of comments are normal at this stage.
Before enquiry, the Commission assesses whether the draft is eligible to become a harmonised standard (HAS assessment).
National standards bodies comment and vote. The draft can fail on the number of countries in favour or on the weighted population in favour.
After a positive enquiry, the draft may be published without a formal vote (CEN-CENELEC Technical Board, October 2025). After a negative enquiry, comments are resolved and a formal vote is held, as happened with EN 18286.
The standard exists and can be bought and applied — but it carries no legal effect yet.
The Commission publishes the reference of the standard in the Official Journal.
Only now does following the standard show compliance with the AI Act requirement it covers.
A finished standard and a standard that gives presumption of conformity are two different things. So far, most of the delay has come before publication: drafts reached enquiry months later than planned, and every enquiry vote on a core draft has been negative. EN 18286 is the first to be published; its citation in the Official Journal is still pending.
steps 01–06 are technical work · steps 07–08 are what makes it law-relevant
Source: Inclusiveness Newsletter, CEN-CENELEC JTC 21 Task Group Inclusiveness (Editions as listed). Summaries by Georg Philip Krog.
A harmonised standard gives a presumption of conformity only when its reference is published in the Official Journal of the EU. Publication by CEN-CENELEC is not enough. The presumption covers only the clauses and conditions listed in the standard's Annex ZA, and it lapses if the citation is withdrawn.
EN 18286
Quality management system for EU AI Act regulatory purposes
Depends on citation
Legal effect status: No Presumption (Published, Not Cited)
Source: Edition 13, Edition 15, Edition 16
prEN 18228
AI risk management
Not yet
Depends on citation
Legal effect status: No Presumption (Draft)
Source: Edition 15, Edition 16
prEN 18282
Cybersecurity specifications for AI systems
Not yet
Depends on citation
Legal effect status: No Presumption (Draft)
Source: Edition 15, Edition 16
prEN 18229-1
AI trustworthiness framework, Part 1: Logging
Not yet
Depends on citation
Legal effect status: No Presumption (Draft)
Source: Edition 15, Edition 16
Source: Inclusiveness Newsletter, CEN-CENELEC JTC 21 Task Group Inclusiveness (Editions as listed). Summaries by Georg Philip Krog.
EN 18286quality management
Published 22 Jul 2026, not yet cited
prEN 18228risk management
Rejected at enquiry, comment resolution November 2026
prEN 18229 (5 parts)trustworthiness framework
Part 1 rejected at enquiry · Part 3 at enquiry · Parts 2, 4, 5 with the Commission
prEN 18282cybersecurity
Rejected at enquiry, revised scope at ballot
prEN 18283bias
Commission comments under review
prEN 18284datasets
Commission comments under review
prEN 18285conformity assessment
Agreed as a mature draft 4 Sep 2026
not started · drafting · enquiry · approved · cited in OJ
Source: Inclusiveness Newsletter, CEN-CENELEC JTC 21 Task Group Inclusiveness (Editions as listed). Summaries by Georg Philip Krog.
Harmonisation roles: Seeks presumption · Referenced (no presumption sought) · Not intended for harmonisation
| Standard | Title | AI Act provision | Working group | Harmonisation role | Current status | Next step | Source |
|---|---|---|---|---|---|---|---|
| EN 18286 | Quality management system for EU AI Act regulatory purposes | Art. 17(1), Art. 11(1), Art. 72 | WG2 | Seeks presumption | Published on 22 July 2026. Not yet cited in the Official Journal. | Citation in the Official Journal | Ed. 13, Ed. 15, Ed. 16 |
| prEN 18228 | AI risk management | Art. 9 | WG2 | Seeks presumption | Rejected at public enquiry (closed 30 July 2026) by a majority of national standards bodies. The Commission's preliminary assessment found it a valid candidate for harmonisation. | Final comment resolution meetings, first week of November 2026 | Ed. 15, Ed. 16 |
| prEN 18229-1 | AI trustworthiness framework, Part 1: Logging | Art. 12 | WG4 | Seeks presumption | Rejected at public enquiry (closed 20 August 2026) on the weighted population criterion. | Reconsideration requests until 25 September; comment resolution 30 September to 2 October 2026 | Ed. 15, Ed. 16 |
| prEN 18229-2 | AI trustworthiness framework, Part 2: Transparency | Art. 13 | WG4 | Seeks presumption | Submitted to the Commission for assessment before public enquiry. | Public enquiry | Ed. 16 |
| prEN 18229-3 | AI trustworthiness framework, Part 3: Human oversight | Art. 14 | WG4 | Seeks presumption | At public enquiry until 22 October 2026. | Enquiry vote | Ed. 16 |
| prEN 18229-4 | AI trustworthiness framework, Part 4: Accuracy | Art. 15 (accuracy) | WG4 | Seeks presumption | Submitted to the Commission for assessment before public enquiry. | Public enquiry | Ed. 16 |
| prEN 18229-5 | AI trustworthiness framework, Part 5: Robustness | Art. 15 (robustness) | WG4 | Seeks presumption | Submitted to the Commission for assessment before public enquiry. | Public enquiry | Ed. 16 |
| prEN 18282 | Cybersecurity specifications for AI systems | Art. 15 (cybersecurity) | WG5 | Seeks presumption | Rejected at public enquiry (closed 30 July 2026) by a majority of national standards bodies. The Commission's preliminary assessment found it a valid candidate for harmonisation. | Ballot on a revised scope until 16 September 2026, then comment resolution | Ed. 15, Ed. 16 |
| prEN 18284 | Quality and governance of datasets in AI systems | Art. 10 | WG3 | Seeks presumption | Sent to the Commission in July 2026 for its pre-enquiry check. Commission comments received in August are under review. | Public enquiry, targeted before the October 2026 plenary | Ed. 13, Ed. 16 |
| EN ISO/IEC 12792 | Transparency taxonomy of AI systems | Art. 13 | — | Referenced (no presumption sought) | Published (EN ISO/IEC 12792:2025). | — | Ed. 13, Ed. 14 |
| ISO/IEC 23282 | Evaluation methods for accurate NLP systems | Art. 15 (accuracy) | WG3 | Referenced (no presumption sought) | At DIS ballot until 25 November 2026. | Ballot result | Ed. 13, Ed. 16 |
| ISO/IEC 24970 | AI system logging | Art. 12 | WG3 | Referenced (no presumption sought) | At ballot until 23 October 2026. | Ballot result | Ed. 13, Ed. 16 |
| prEN 18281 | Evaluation methods for accurate computer vision systems | Art. 15 (accuracy) | WG3 | Referenced (no presumption sought) | Rejected at public enquiry (closed 11 June 2026). | — | Ed. 13, Ed. 15 |
| prEN 18283 | Concepts, measures and requirements for managing bias in AI systems | Art. 10(2)(f)–(g) | WG3 | Referenced (no presumption sought) | Sent to the Commission in July 2026 for its pre-enquiry check. Commission comments received in August are under review. | Public enquiry, targeted before the October 2026 plenary | Ed. 13, Ed. 16 |
| prEN 18285 | AI conformity assessment framework | Art. 43, Annexes VI and VII | WG2 | Referenced (no presumption sought) | Agreed as a mature draft on 4 September 2026, to be sent to the Commission for review. Revised scope at ballot until 21 September 2026. | Commission review | Ed. 13, Ed. 16 |
| EN ISO/IEC 42001 | AI management system | — | WG2 | Not intended for harmonisation | Approved for direct adoption as a European standard. Not intended to be harmonised under the AI Act. | — | Ed. 13, Ed. 14 |
| prCEN/CLC/TR 18347 | Overview and architecture of standards in support of the EU AI Act | — | WG1 | Not intended for harmonisation | Second working draft open for comment until 1 October 2026. | — | Ed. 15, Ed. 16 |
| prEN ISO/IEC 42006 | Requirements for bodies auditing and certifying AI management systems | — | WG1 | Not intended for harmonisation | At enquiry for direct adoption until 29 October 2026. | Enquiry vote | Ed. 15, Ed. 16 |
Quality management system for EU AI Act regulatory purposes
Source: Ed. 13, Ed. 15, Ed. 16
AI risk management
Source: Ed. 15, Ed. 16
AI trustworthiness framework, Part 1: Logging
Source: Ed. 15, Ed. 16
AI trustworthiness framework, Part 2: Transparency
Source: Ed. 16
AI trustworthiness framework, Part 3: Human oversight
Source: Ed. 16
AI trustworthiness framework, Part 4: Accuracy
Source: Ed. 16
AI trustworthiness framework, Part 5: Robustness
Source: Ed. 16
Cybersecurity specifications for AI systems
Source: Ed. 15, Ed. 16
Quality and governance of datasets in AI systems
Source: Ed. 13, Ed. 16
Transparency taxonomy of AI systems
Source: Ed. 13, Ed. 14
Evaluation methods for accurate NLP systems
Source: Ed. 13, Ed. 16
AI system logging
Source: Ed. 13, Ed. 16
Evaluation methods for accurate computer vision systems
Source: Ed. 13, Ed. 15
Concepts, measures and requirements for managing bias in AI systems
Source: Ed. 13, Ed. 16
AI conformity assessment framework
Source: Ed. 13, Ed. 16
AI management system
Source: Ed. 13, Ed. 14
Overview and architecture of standards in support of the EU AI Act
Source: Ed. 15, Ed. 16
Requirements for bodies auditing and certifying AI management systems
Source: Ed. 15, Ed. 16
Source: Inclusiveness Newsletter, CEN-CENELEC JTC 21 Task Group Inclusiveness (Editions as listed). Summaries by Georg Philip Krog.
At public enquiry and formal vote, a draft needs enough national standards bodies in favour and enough weighted votes, which are based on population, in favour. Failing either criterion means the draft is rejected and its comments must be resolved before it can move on.
Since January 2026, 5 of 6 recorded votes rejected a draft. Only EN 18286 has passed, at its second attempt.
prEN 18229-1
AI trustworthiness framework, Part 1: Logging
Public enquiry · closed 2026-08-20
443 comments
Comment resolution, 30 September to 2 October 2026
Source: Edition 16
prEN 18228
AI risk management
Public enquiry · closed 2026-07-30
Not met
Not reported
Comment resolution, first week of November 2026
Source: Edition 16
prEN 18282
Cybersecurity specifications for AI systems
Public enquiry · closed 2026-07-30
Not met
Not reported
Revised scope at ballot, then comment resolution
Source: Edition 16
EN 18286
Quality management system for EU AI Act regulatory purposes
Formal vote · closed 2026
Published on 22 July 2026
Source: Edition 15, Edition 16
prEN 18281
Evaluation methods for accurate computer vision systems
Public enquiry · closed 2026-06-11
Not reported
Not reported
Source: Edition 15
EN 18286
Quality management system for EU AI Act regulatory purposes
Public enquiry · closed January 2026
Not met
Not met
1,288 comments
Reconsideration requests handled in London, 2–6 March 2026
Source: Edition 13
Source: Inclusiveness Newsletter, CEN-CENELEC JTC 21 Task Group Inclusiveness (Editions as listed). Summaries by Georg Philip Krog.
Trustworthiness, first draft (comments reported February 2025) prEN 18229
over 2,000 comments
Restructured afterwards
Risk management, committee draft prEN 18228
1,300 comments
Moved on to enquiry
Quality management, committee draft prEN 18286
1,100 comments
Moved on to enquiry
Quality management, public enquiry, January 2026 EN 18286
1,288 comments
Failed on both country count and weighted population
Logging (prEN 18229-1), public enquiry, August 2026 prEN 18229-1
443 comments
Rejected on weighted population
number of comments received on each draft · red marks a vote that did not pass
Source: Inclusiveness Newsletter, CEN-CENELEC JTC 21 Task Group Inclusiveness (Editions as listed). Summaries by Georg Philip Krog.
Each Working group drafts specific deliverables. Horizontal work, such as inclusiveness or technical coherence, sits in a Task Group.
Strategic advisory
Coherence of the work programme, terminology, inclusiveness, and the overview of standards supporting the AI Act.
Main deliverables
prCEN/CLC/TR 18347
Operational aspects
Quality management, risk management and conformity assessment.
Main deliverables
EN 18286 · prEN 18228 · prEN 18285
Engineering aspects
Datasets, bias, logging (with ISO/IEC), natural language processing and computer vision.
Main deliverables
prEN 18284 · prEN 18283 · ISO/IEC 24970
Foundational and societal aspects
Trustworthiness framework (logging, transparency, human oversight, accuracy, robustness), ethics, environmental impact and fundamental rights.
Main deliverables
prEN 18229-1 to 18229-5
Cybersecurity
Security of AI systems; formed later than the other four groups.
Main deliverables
prEN 18282
five working groups · one shared deadline
Source: Inclusiveness Newsletter, CEN-CENELEC JTC 21 Task Group Inclusiveness (Editions as listed). Summaries by Georg Philip Krog.
1 Aug 2024
AI Act enters into force
2 Feb 2025
Prohibited practices and AI literacy apply
2 Aug 2025
GPAI model obligations, governance and penalties apply
2 Aug 2026
General application, including Article 50 transparency duties
2 Dec 2026
Article 50(2) marking obligations apply to AI systems already on the market
28 Feb 2027
Standardisation request M/613 expires
2 Aug 2027
Deadline for general-purpose AI models placed on the market before 2 August 2025
2 Dec 2027
Annex III high-risk duties and Article 27 FRIA apply (Digital Omnibus)
2 Aug 2028
Annex I high-risk duties apply (Digital Omnibus)
3 months to article 50(2) marking obligations apply to ai systems already on the market
You are here · 2026-09-15
1 Aug 2024 · in force
AI Act enters into force
2 Feb 2025 · in force
Prohibited practices and AI literacy apply
2 Aug 2025 · in force
GPAI model obligations, governance and penalties apply
2 Aug 2026 · in force
General application, including Article 50 transparency duties
2 Dec 2026 · next
Article 50(2) marking obligations apply to AI systems already on the market
28 Feb 2027
Standardisation request M/613 expires
2 Aug 2027
Deadline for general-purpose AI models placed on the market before 2 August 2025
2 Dec 2027
Annex III high-risk duties and Article 27 FRIA apply (Digital Omnibus)
2 Aug 2028
Annex I high-risk duties apply (Digital Omnibus)
3 months to article 50(2) marking obligations apply to ai systems already on the market
filled = already in force · hollow = still ahead · flag = date of this issue
Source: Inclusiveness Newsletter, CEN-CENELEC JTC 21 Task Group Inclusiveness (Editions as listed). Summaries by Georg Philip Krog.
This page summarises public information and is not an official CEN-CENELEC or Commission source. Verify against the CEN-CENELEC work programme and the Official Journal before relying on it.