Do you need a notified body for your high-risk AI system?

"Will we need an external auditor for this?" is one of the first questions a company asks when it discovers its AI system might be high-risk. It is the question that decides whether compliance is a documentation project or a procurement project with a queue in front of it.

The answer is better than most people fear, and worse than it looks in one specific corner. For seven of the eight Annex III areas the EU AI Act lets you assess your own system. For one Annex III category a notified body can be mandatory — and as of 27 September 2026 the Commission's own register listed none designated under the AI Act at all.

This is an information service to help you plan, not legal advice. To see which obligations attach to your own AI use, start with the obligation check.

Three routes under Article 43, and one of them never involves a third party

Article 43 sorts every high-risk AI system it applies to into one of three procedures. It does not apply to AI in products under Annex I Section B (vehicles, aviation, rail, marine equipment and, since the Digital Omnibus, machinery). For those, Article 2(2) applies only Article 6(1), Article 60a and Articles 102 to 112 of the AI Act (plus the sandbox Articles 57 to 59, but only in so far as the AI requirements have been integrated into that product legislation), so any conformity assessment, and any third-party involvement, comes from the product's own legislation rather than from a row below. Everyone else: find your row first; almost everything else follows from it.

Your system	Provision	Procedure	Notified body?
Annex III points 2–8 — education, employment, essential public and private services, law enforcement, migration, justice, critical infrastructure	Article 43(2)	Internal control, Annex VI	No — the text says the procedure "does not provide for the involvement of a notified body"
Annex III point 1 — biometrics	Article 43(1)	Annex VI or Annex VII, and Annex VII alone in the cases listed below	Sometimes, and sometimes mandatorily
Annex I Section A products — the product route under Article 6(1)	Article 43(3)	The sectoral procedure under that product legislation	Whatever that legislation requires, except that a standards-based opt-out from third-party assessment is open only if AI Act harmonised standards or common specifications covering Section 2 have also been applied

If your system fits two rows, because it is covered by Annex I Section A legislation and also falls in an Annex III category, the Omnibus added a tie-breaker to the end of Article 43(3): the provider "shall follow the relevant conformity assessment procedure as required pursuant to the relevant Union harmonisation legislation listed in Section A of Annex I". The product row wins.

Article 43(2), in full:

"For high-risk AI systems referred to in points 2 to 8 of Annex III, providers shall follow the conformity assessment procedure based on internal control as referred to in Annex VI, which does not provide for the involvement of a notified body."

That is the sentence most companies are looking for. If your high-risk AI is a recruitment screener, a credit-scoring engine, an education tool or a system used in essential services, no third party has to certify it. You build the technical documentation, run the quality management system, draw up the EU declaration of conformity under Article 47 and affix the CE marking under Article 48 yourself. It is a serious amount of work. It is not a gated audit.

If you are still working out whether you are high-risk at all, the Article 6 test walks through both doors — and if you are in an Annex III area but doing something narrow, check the Article 6(3) derogation before any of this applies to you.

The biometrics exception, and why it is the tight spot

Annex III point 1 is the one category where Article 43(1) can force a notified body. The new Annex XIV spells out what sits inside it, in the codes a notified body's designation is scoped by:

Annex XIV code	Type of AI system
AIB 0201	Remote biometric identification systems
AIB 0202	Biometric categorisation AI systems
AIB 0203	Emotion recognition AI systems

One boundary: in the workplace and in education institutions, emotion recognition does not get as far as conformity assessment. Article 5(1)(f) prohibits placing on the market, putting into service for this specific purpose, or using AI systems "to infer emotions of a natural person in the areas of workplace and education institutions", except where the system is intended for medical or safety reasons. Where that line falls is set out in is AI emotion recognition banned at work?.

For the systems in that table, Article 43(1) gives the provider a choice between Annex VI internal control and Annex VII with a notified body — but only "where … the provider has applied harmonised standards referred to in Article 40, or, where applicable, common specifications referred to in Article 41".

Where that condition is not met, the choice disappears. The second subparagraph is mandatory:

"the provider shall follow the conformity assessment procedure set out in Annex VII where: (a) harmonised standards referred to in Article 40 do not exist, and common specifications referred to in Article 41 are not available; (b) the provider has not applied, or has applied only part of, the harmonised standard; (c) the common specifications referred to in point (a) exist, but the provider has not applied them; (d) one or more of the harmonised standards referred to in point (a) has been published with a restriction, and only on the part of the standard that was restricted."

Now read that against where the standards actually are. Article 40 harmonised standards are standards whose references have been published in the Official Journal. EN 18286:2026 — the first AI Act harmonised standard — is published but, as at 27 September 2026, has no Official Journal reference. Until it is cited, it is not an Article 40 harmonised standard and confers no presumption of conformity. See what EN 18286 does and does not give you.

So for an Annex III point 1 provider today, condition (a) is the live one, and Annex VII is not a choice. Two exceptions change who does the assessing. Where the system is intended to be put into service by law enforcement, immigration or asylum authorities or by EU institutions, bodies, offices or agencies, Article 43(1) says the market surveillance authority referred to in Article 74(8) or (9) "shall act as a notified body". And where the system falls under the AI Office's exclusive supervision under Article 75(1) — that is, an AI system based on a general-purpose AI model where the model and the system come from the same provider or undertaking (subject to the exceptions that paragraph lists), or one that constitutes or is integrated into a designated very large online platform or search engine — Article 75(1e), inserted by the Digital Omnibus, makes the AI Office "responsible for conformity assessments and tests" of high-risk systems "subject to a third-party conformity assessment pursuant to Article 43". The provider applies to the Commission, and the Commission entrusts the work to notified bodies designated under the AI Act, which act on its behalf — the same designated bodies whose register, below, is empty.

What the Commission's own register shows

Article 35(2) puts the count in the public domain:

"The Commission shall make publicly available the list of the bodies notified under this Regulation, including their identification numbers and the activities for which they have been notified. The Commission shall ensure that the list is kept up to date."

That list lives in the Commission's Single Market Compliance Space database, in its notified-bodies section (NANDO). Filtered to Regulation (EU) 2024/1689 with notification status "Active", checked on 27 September 2026, it returns no results.

For scale, the same database on the same day:

Legislation	Active notified bodies
Regulation (EU) 2017/745 on medical devices	52
Regulation (EU) 2023/1230 on machinery	41
Regulation (EU) 2024/2847 (Cyber Resilience Act)	0
Regulation (EU) 2024/1689 (AI Act)	0

The machinery row is the one to sit with. Regulation (EU) 2023/1230 as a whole does not apply until 20 January 2027, but its notified-body articles (Articles 26 to 42) have applied since 20 January 2024 under its Article 54, as corrected, and it has 41 designated bodies. The AI Act's notified-body chapter — Chapter III Section 4, Articles 28 to 39 — has applied since 2 August 2025 under Article 113(b), and the register is empty.

We are reporting what the Commission's list shows on a given date, not asserting that no body anywhere is in the middle of an application. Designation runs through a notifying authority under Article 29, then a notification to the Commission and Member States under Article 30, and a body "may perform the activities of a notified body only where no objections are raised … within two weeks" of an accredited notification or "within two months" of an unaccredited one. Bodies in that pipeline would not yet appear.

If you build Annex I Section A products, the Omnibus changed your route

Article 43(3) was rewritten by the Digital Omnibus, Regulation (EU) 2026/1744, and the rewrite cuts three ways.

It keeps your existing notified body in play, but on a clock. That part is not new: the 2024 text already said bodies notified under the Section A legislation were "entitled to control the conformity" of high-risk AI systems with Section 2. What the Omnibus adds is the clock. Recital 18 describes the power as applying "for 18 months from 27 July 2026", and the same subparagraph now ends with the designation deadline below. Bodies already notified under the Annex I Section A product legislation "shall have the power to assess the conformity of high-risk AI systems with the requirements set out in Section 2 of this Chapter", provided their compliance with Article 31(4), (5), (10) and (11) was assessed in the sectoral notification procedure. Your medical-device notified body, for example, does not have to wait for a separate AI Act designation to look at the AI requirements. Machinery is no longer on this route: the Omnibus deleted Directive 2006/42/EC from Annex I Section A and added Regulation (EU) 2023/1230 on machinery to Section B, and under Article 2(2) Article 43 does not apply to Section B products at all. The Omnibus instead amended the Machinery Regulation itself: its Article 8 now requires the Commission to add health and safety requirements for high-risk AI systems to that Regulation's Annex III by delegated act, and those delegated acts "shall apply by 2 August 2028".

It says the AI alone does not drag the product into third-party assessment. This sentence is new:

"The manufacturers of such products are not required to choose a conformity assessment procedure involving third-party conformity assessment only because the product includes a high-risk AI system as a safety component, if this is not required by the Union harmonisation legislation listed in Section A of Annex I."

That answers a genuine fear among product manufacturers: that adding an AI component would, by itself, turn a self-assessed product into an audited one. But read it together with the first sentence of the same subparagraph, which the Omnibus kept in substance. Where the product legislation lets a manufacturer avoid a third party because it has applied harmonised standards, it may use that option "only if it has also applied harmonised standards or, where applicable, common specifications referred to in Article 41, covering all requirements set out in Section 2 of this Chapter". Until AI Act harmonised standards or common specifications covering Section 2 exist, that condition is hard to meet, and the text does not say how it fits with the new sentences that follow it. For the wider picture of that route, see the safety-component question.

It adds work, and a new deadline. The amended paragraph also requires that "Assessment of the quality management system set out in Article 17 shall also be undertaken", with points 3, 4.3, 4.4, 4.5, the fifth paragraph of point 4.6 and point 5 of Annex VII applying — where the original text applied only points 4.3, 4.4, 4.5 and the fifth paragraph of 4.6. And it sets a date:

"Without prejudice to Article 28, such notified bodies which have been notified under the Union harmonisation legislation in Section A of Annex I, shall apply for designation in accordance with Section 4 of this Chapter by 28 January 2028."

28 January 2028 is a hard deadline created by the Omnibus, and it is missing from the Commission's own reference page for Article 43, which, as at 29 September 2026, still displays the pre-Omnibus text under the notice: "This provision has been amended by the Digital Omnibus on AI. The text displayed on this page has not yet been updated to reflect those amendments." If you rely on a sectoral notified body, that is the date by which it must apply for AI Act designation — and it is a question worth putting to your body now rather than in 2027.

Annex XIV: the new annex that scopes a designation

Annex XIV did not exist in the 2024 text. The Omnibus added it, and its purpose is narrow but consequential: a notified body is not designated "for the AI Act", it is designated for specific types of AI system.

"Only conformity assessment bodies that have been designated in accordance with this Regulation may carry out conformity assessments and only for the activities related to the types of AI systems concerned. The list of codes, categories, and corresponding types of AI systems sets the scope of the designation of conformity assessment bodies notified under Article 30."

Point 4 of the annex makes it operational: conformity assessment bodies "shall use the lists of codes, categories and corresponding types of AI systems set out in this Annex when specifying the types of AI systems in the application for designation referred to in Article 29". Amended Article 30(2) then requires notifying authorities to notify "based on the list of codes, categories, and corresponding types of AI systems referred to in Annex XIV, and using the electronic notification tool developed and managed by the Commission".

Alongside the product codes (AIP 0102 to AIP 0112, tracking points 2 to 12 of Annex I Section A) and the biometric codes above, the annex carries a set of technology codes:

Code	Category
AIH 0101	Symbolic AI, expert and knowledge-based systems, search and optimisation
AIH 0201–0205	Machine learning, by data type — structured, signal and audio, text, image and video, and systems that learn from their environment
AIH 0301	"generative AI systems, including AI systems based on general-purpose AI models"
AIH 0401	"AI systems based on other emerging AI technologies not covered by other codes, including Agentic AI"

AIH 0401 is worth noting for its own sake: the phrase "Agentic AI" now appears in the binding text of the EU AI Act, put there by the Digital Omnibus. It is a designation-scope label, not a new obligation — there is no separate agentic-AI regime — but it tells you the EU legislator expects agentic systems to reach conformity assessment, and it is the first foothold the Act gives that category.

The Commission may amend Annex XIV by delegated act under Article 97, adding, withdrawing or moving codes "in light of technical progress, advances in knowledge or new scientific evidence". Expect the list to grow.

Designation got easier for the bodies themselves

Two new paragraphs in Article 28 are aimed squarely at the bottleneck above. Article 28(8) requires notifying authorities to give a body applying under both the AI Act and Annex I Section A legislation the option of "a single application" and "a unified assessment procedure", where the sectoral legislation provides for it — and makes it available to bodies already designated sectorally. It goes further:

"A conformity assessment body that is designated pursuant to more than one piece of Union harmonisation legislation listed in Section A of Annex I shall have to apply only once to be designated pursuant to this Regulation. A designation pursuant to this Regulation shall be applicable for all Union harmonisation legislation listed in Section A of Annex I for which the conformity assessment body is designated."

Article 28(9) settles which authority runs that single procedure. Whether this moves the register is an empirical question, and one we will keep checking.

The dates

Provision	Where it sits	Applies from
Notifying authorities and notified bodies (Articles 28–39)	Chapter III, Section 4	2 August 2025
Conformity assessment, CE marking, registration (Articles 40–49)	Chapter III, Section 5	Not named in Article 113; general rule, 2 August 2026
Classification, requirements, provider and deployer obligations (Articles 6–27, except Article 6(5))	Chapter III, Sections 1–3	2 December 2027 (Annex III) / 2 August 2028 (Annex I)
Existing Annex I Section A notified bodies to apply for AI Act designation	Article 43(3)	by 28 January 2028

Two notes on that table. Article 113 as amended by the Digital Omnibus defers "Chapter III, Sections 1, 2, and 3" only — Section 5, which contains Article 43, is not named in it, so on the face of the text it runs on the Regulation's general date. Many summaries treat the whole of Chapter III as deferred; that is a reasonable practical reading and it is not what Article 113 says. We flag it as unresolved rather than picking a side, and we have found no Commission guidance on the sequencing: the AI Act Service Desk's implementation timeline and FAQ do not address it.

Either way, the substantive requirements a conformity assessment tests — Articles 8 to 15, in Section 2 — arrive on 2 December 2027 for Annex III systems. That is the date that decides when the empty register starts to matter. See the deadline that moved to 2027 and the full timeline.

What to do now

1. Find your row in the first table. If you are Annex III points 2–8, and your system is not also covered by Annex I Section A product legislation, the notified-body question is closed: Article 43(2), internal control, no third party. Put the effort into Articles 9 to 15 and the Article 17 quality management system instead. If your AI sits in an Annex I Section B product, such as a vehicle or, since the Omnibus, machinery, you have no row: Article 43 does not apply, and the question is answered under that product's own legislation.
2. If you are Annex III point 1, plan for Annex VII. With no Article 40 harmonised standard cited for the AI Act, the mandatory branch of Article 43(1) is the live one. Build the Annex VII documentation set on the assumption that an external body will read it. If your system is built on a general-purpose AI model developed by you or by your group, or constitutes or is integrated into a designated very large online platform or search engine, check Article 75(1) and its exceptions first: where it applies, Article 75(1e) has you apply to the Commission, which entrusts the assessment to a designated notified body acting on its behalf.
3. If you build Annex I Section A products, talk to your existing notified body about 28 January 2028. Ask whether it intends to apply for AI Act designation, and for which Annex XIV codes. Recital 18 frames the existing power as lasting 18 months from 27 July 2026, so do not assume that a body which has not applied can keep assessing your AI requirements once those 18 months run out in late January 2028.
4. Match your system to Annex XIV codes now. When bodies do appear, they will be designated per code, not for the Act as a whole. Knowing that you may need, for example, AIB 0202 and AIH 0204 turns a search into a filter.
5. Do not treat Article 46 as a plan. A market surveillance authority may authorise placing a specific high-risk system on the market by derogation from Article 43, but only "for exceptional reasons of public security or the protection of life and health of persons, environmental protection or the protection of key industrial and infrastructural assets", and only for a limited period while the conformity assessment is carried out. It is not a commercial route around an empty register.
6. Watch two things. The Official Journal, for Commission implementing decisions citing AI Act harmonised standards; and the Commission's notified-body list, for the first AI Act designation. On the first, note that EN 18286 is a quality management system standard, while the Article 43(1) choice depends on harmonised standards applied to the Section 2 requirements. Citing EN 18286 alone would not obviously open that choice for biometrics providers.

If you would rather be told when those two things happen than check them yourself, join the waitlist for the weekly email of verified AI Act changes we are building.

---

Primary sources: Regulation (EU) 2024/1689, read in the consolidated version as at 27 July 2026, as amended by Regulation (EU) 2026/1744; notified-body counts from the Commission's Single Market Compliance Space database, checked 27 September 2026.

Reglog is an information service that tracks changes to the EU AI Act and tells you which ones touch your obligations. It is not legal advice, and it does not create a lawyer–client relationship. For a decision with consequences, take qualified advice on your specific facts.