Guide · EU AI Act
Is your AI system high-risk? EU AI Act Article 6 and Annex III, read in full
An AI system is high-risk under the EU AI Act by one of two routes in Article 6, as a safety component of a product regulated under Annex I that needs third-party conformity assessment, or as a use listed in one of the eight areas of Annex III. An Annex III system escapes the class only under the Article 6(3) exception, and never if it profiles people.
By derogation from paragraph 2, an AI system referred to in Annex III shall not be considered to be high-risk where it does not pose a significant risk of harm to the health, safety or fundamental rights of natural persons, including by not materially influencing the outcome of decision making.
The figure behind it
Annex III lists 8 areas of high-risk use, and after Regulation (EU) 2026/1744 the high-risk duties for those systems apply from 2 December 2027. Source: Regulation (EU) 2024/1689 as amended by Regulation (EU) 2026/1744, Annex III (8 paragraphs) and Article 113, as held by Rekvira (consolidated text of 27 July 2026, corpus of 24 September 2026)
The wording below is the Act as it stands today, Regulation (EU) 2024/1689 as amended by Regulation (EU) 2026/1744, returned by Rekvira's tools on 24 September 2026. The July 2026 amendment left Annex III and Article 6(2) to (8) as they were, added three paragraphs to Article 6 and moved the application dates; the tools mark each amended paragraph with the act that changed it.
What are the two routes into the high-risk class?
Article 6 opens with the product route. Both conditions have to hold:
"(a) the AI system is intended to be used as a safety component of a product, or the AI system is itself a product, covered by the Union harmonisation legislation listed in Annex I; (b) the product whose safety component pursuant to point (a) is the AI system, or the AI system itself as a product, is required to undergo a third-party conformity assessment" (Article 6(1))
The second route is a list: "In addition to the high-risk AI systems referred to in paragraph 1, AI systems referred to in Annex III shall be considered to be high-risk" (Article 6(2)). For most companies that do not make regulated products, Annex III is the route that decides the question.
Regulation (EU) 2026/1744 added three paragraphs to Article 6 that narrow the product route. The first reads: "AI systems that are solely used for non-safety related aspects of user assistance, performance optimisation, service efficiency, automation or convenience or quality control shall not qualify as safety components" (Article 6(1a), as inserted by Article 1(6) of that Regulation).
What does Annex III list?
Eight areas. The Annex names specific uses inside each, so the area heading alone does not make a system high-risk.
| # | Area | Uses the Annex names (selection, verbatim) |
|---|---|---|
| 1 | Biometrics | "remote biometric identification systems"; "emotion recognition" |
| 2 | Critical infrastructure | "safety components in the management and operation of critical digital infrastructure, road traffic, or in the supply of water, gas, heating or electricity" |
| 3 | Education and vocational training | "to determine access or admission"; "to evaluate learning outcomes" |
| 4 | Employment and workers' management | "to analyse and filter job applications, and to evaluate candidates" |
| 5 | Essential private and public services | "to evaluate the creditworthiness of natural persons or establish their credit score"; "risk assessment and pricing in relation to natural persons in the case of life and health insurance" |
| 6 | Law enforcement | "to evaluate the reliability of evidence in the course of the investigation or prosecution of criminal offences" |
| 7 | Migration, asylum and border control | "for the examination of applications for asylum, visa or residence permits" |
| 8 | Administration of justice and democratic processes | "to assist a judicial authority in researching and interpreting facts and the law" |
read_unit on 24 September 2026 (8 paragraph pinpoints). The Annex has more uses than shown; read it whole before you rule a system out.Two carve-outs sit inside the list and are easy to miss. Biometric verification whose "sole purpose" is to confirm a person is who they claim to be is excluded from point 1(a). Credit scoring "for the purpose of detecting financial fraud" is excluded from point 5(b).
When is an Annex III system not high-risk?
Article 6(3) is the exception, and it has two tests that apply together. The system must not pose a significant risk of harm, including by not materially influencing the outcome of a decision, and one of four conditions must be met:
4
conditions in Article 6(3), of which one must apply
1
override: profiling of natural persons is always high-risk
8
Annex III areas the exception can apply to
read_unit on 24 September 2026.| Condition | Article 6(3) wording |
|---|---|
| (a) | "the AI system is intended to perform a narrow procedural task" |
| (b) | "the AI system is intended to improve the result of a previously completed human activity" |
| (c) | "the AI system is intended to detect decision-making patterns or deviations from prior decision-making patterns and is not meant to replace or influence the previously completed human assessment, without proper human review" |
| (d) | "the AI system is intended to perform a preparatory task to an assessment relevant for the purposes of the use cases listed in Annex III" |
Then the override: "an AI system referred to in Annex III shall always be considered to be high-risk where the AI system performs profiling of natural persons." A CV-ranking tool that builds a profile of each candidate cannot use the exception, whichever of the four conditions it might otherwise meet.
What must you do if you rely on the exception?
Write it down first. "A provider who considers that an AI system referred to in Annex III is not high-risk shall document its assessment before that system is placed on the market or put into service" (Article 6(4)). The same paragraph keeps the system inside the EU database: the provider "shall be subject to the registration obligation set out in Article 49(2)". The July 2026 amendment simplified what that registration contains, not whether it happens.
A good test before you file the assessment: can you name the condition (a) to (d) you rely on, quote it, and explain why the system does not materially influence the decision? If any of the three is missing, you do not have an exception.
Which duties follow, and from when?
For a high-risk system, the obligations attach to the role. Rekvira's heading lookup returns Article 16 for providers and Article 26 for deployers, with related articles such as Article 22 (authorised representatives), Article 72 (post-market monitoring) and Article 80 (the procedure for systems a provider classified as not high-risk). The full split is in EU AI Act compliance: the obligations by role.
The dates changed in July 2026. The Act as published set 2 August 2026 as the general date and 2 August 2027 for Article 6(1). Regulation (EU) 2026/1744 replaced Article 113, third paragraph, point (c):
"Chapter III, Sections 1, 2, and 3, with the exception of Article 6(5), shall apply from: (i) 2 December 2027 as regards AI systems classified as high-risk pursuant to Article 6(2) and Annex III; and (ii) 2 August 2028 as regards AI systems classified as high-risk pursuant to Article 6(1) and Annex I" (Article 113, as replaced by Regulation (EU) 2026/1744, Article 1(40))
Chapters I and II apply: definitions, AI literacy, prohibited practices (Article 113(a); the prohibitions added in 2026 apply from 2 December 2026)
General-purpose AI models, governance and penalties apply (Article 113(b))
High-risk duties apply to Annex III systems (Article 113(c)(i), as amended in 2026)
High-risk duties apply to Annex I product systems (Article 113(c)(ii), as amended in 2026)
Recorded run
The question:Is our AI system high-risk under the EU AI Act, and from when do the high-risk duties apply?
get_workflow(workflow_id="high-risk-triage")Load the triage playbook and its stop rules.
read_unit(regulation="eu-ai-act", kind="article", number="6")The classification rules, whole.
Article 6 whole, 11 paragraph pinpoints (1a to 1c marked as inserted by Regulation (EU) 2026/1744), including "shall always be considered to be high-risk where the AI system performs profiling of natural persons"
read_unit(regulation="eu-ai-act", kind="annex", number="III")The eight Annex III areas, whole.
Annex III in 8 pinpointed areas, from "Biometrics" to "Administration of justice and democratic processes"
verify_citation(regulation="eu-ai-act", citation="Article 6(3)", quote="shall not be considered to be high-risk where it does not pose a significant risk of harm to the health, safety or fundamental rights of natural persons")Check the exception wording before quoting it.
Quote found for the Article 6(3) exception wording at the top of this page.
lookup_obligations(regulation="eu-ai-act", role="provider", system_class="high-risk")Which article headings name the provider of a high-risk system.
Heading lookup: "Obligations of providers of high-risk AI systems" (Article 16), with "Post-market monitoring by providers" among the related articles.
lookup_obligations(regulation="eu-ai-act", role="deployer", system_class="high-risk")The same for the deployer.
Heading lookup: "Obligations of deployers of high-risk AI systems" (Article 26).
read_unit(regulation="eu-ai-act", kind="article", number="113")The application dates as the corpus holds them.
Article 113 as amended: "2 December 2027 as regards AI systems classified as high-risk pursuant to Article 6(2) and Annex III".
Real calls to Rekvira’s production server, captured 24 September 2026. Corpus as of 24 September 2026.
Limits
Rekvira holds the AI Act as amended by Regulation (EU) 2026/1744 and answers with the wording in force on the date asked; the 2024 wording of an amended provision is still available for an earlier date, marked as superseded. Whether a system is high-risk is a determination for the provider and its advisers. Rekvira gives the text and the pinpoint, not the conclusion.
Do this in your assistant
Use Rekvira to classify an AI system under the EU AI Act. Read Article 6 and Annex III in full.
For the system I describe, say which Annex III point, if any, it falls under and quote it; then
check each Article 6(3) condition and the profiling override, quoting the words you rely on.Paste this into Claude, ChatGPT, Cursor or any MCP client connected to https://mcp.rekvira.com/mcp. No account is needed to start.Install steps for each client
Start with the keyless trial.
No account, no API key, and free during the beta.
https://mcp.rekvira.com/mcp