High-risk AI checklist: EU AI Act classification in 10 steps

An AI system is high-risk under the EU AI Act if it is a safety component of, or itself, a product that needs third-party conformity assessment under the EU laws in Annex I, or if it is used for one of the purposes listed in Annex III, such as recruitment, credit scoring or education. This high-risk AI checklist takes you through those tests and the Article 6(3) exemption in ten steps. After the Digital Omnibus, the obligations apply from 2 December 2027 for Annex III systems and 2 August 2028 for Annex I products.

Key takeaways

• An AI system is high-risk under the EU AI Act by one of two routes: it is a safety component of (or itself) a product needing third-party conformity assessment under Annex I legislation, or it is used in one of the eight Annex III areas.
• Annex III systems can escape high-risk status under Article 6(3) if they perform narrow procedural, improving, pattern-detecting or preparatory tasks, but never if they profile people.
• A provider relying on the Article 6(3) exemption must document the assessment before launch and register the system.
• High-risk obligations apply from 2 December 2027 for Annex III systems and 2 August 2028 for Annex I products, after the Digital Omnibus.
• Commission guidelines on classification were still in draft at the time of writing, so record your reasoning and revisit it when they are final.

Why does high-risk classification matter so much?

Because almost all of the AI Act's cost sits in the high-risk tier. Providers of high-risk systems must, under Article 16 of the AI Act, meet the requirements on risk management, data, documentation, logging, transparency to deployers, human oversight, accuracy and robustness; run a quality management system; complete a conformity assessment; issue an EU declaration of conformity; affix the CE marking; and register the system. Deployers have their own duties under Article 26, including human oversight by trained staff and keeping logs for at least six months.

The dates come from the consolidated Article 113 as amended by Regulation (EU) 2026/1744.

What is the high-risk AI checklist?

Work through these ten questions in order and write down your answer and reasoning for each. The wording follows the consolidated Article 6 and Annex III.

1. Is it an AI system at all?

The Act defines an AI system as a machine-based system that operates with some autonomy and infers from its inputs how to generate outputs such as predictions, content, recommendations or decisions. Purely rule-based software written by people usually falls outside. Record: a one-paragraph description of how the system produces its output.

2. Is it a prohibited practice?

Check Article 5 before anything else. Banned uses include harmful manipulation, exploiting vulnerabilities, social scoring, emotion recognition in workplaces and schools (with medical or safety exceptions) and untargeted scraping of facial images. The Omnibus adds bans on generating non-consensual intimate material and child sexual abuse material from 2 December 2026. A prohibited system cannot be made compliant; it has to be redesigned.

3. Is it a product, or a safety component of a product, under Annex I legislation?

Annex I lists EU product safety laws; the Commission gives lifts and toys as examples, and machinery is also covered. If your AI is the product, or is intended as its safety component, continue to question 4. If not, skip to question 6.

4. Is it really a safety component?

The Omnibus narrowed this. AI used solely for non-safety aspects such as user assistance, performance optimisation, efficiency, automation, convenience or quality control is not a safety component, unless its failure or malfunction would endanger health and safety, in which case it is.

5. Does the product need third-party conformity assessment?

Article 6(1)(b) requires the product to undergo third-party conformity assessment under the Annex I law. Under new paragraph 1c, assessment needed only for non-health and safety risks, such as radio spectrum or electromagnetic interference, does not count. If questions 3 to 5 are all yes: the system is high-risk, from 2 August 2028.

6. Is it used in one of the eight Annex III areas?

The areas are: biometrics; critical infrastructure; education and vocational training; employment and worker management; access to essential private and public services; law enforcement; migration, asylum and border control; and administration of justice and democratic processes. For most private companies, the relevant entries are:

Annex III area	Examples that are listed
4. Employment	Recruitment and selection, filtering applications, evaluating candidates; decisions on promotion, termination, task allocation; monitoring and evaluating workers
5. Essential services	Creditworthiness and credit scoring (fraud detection excluded); risk assessment and pricing for life and health insurance; emergency call triage
3. Education	Admission decisions, evaluating learning outcomes, monitoring students during tests
1. Biometrics	Remote biometric identification, biometric categorisation by sensitive attributes, emotion recognition (pure identity verification excluded)

Read the full wording: the use, not the technology, decides.

7. Does the system profile natural persons?

If an Annex III system profiles people, it "shall always be considered to be high-risk". Stop here; the exemption in question 8 is not available.

8. Does an Article 6(3) condition apply?

An Annex III system is not high-risk if it does not pose a significant risk of harm and is intended to: (a) perform a narrow procedural task; (b) improve the result of a previously completed human activity; (c) detect decision-making patterns or deviations without replacing or influencing the human assessment without proper review; or (d) perform a preparatory task to an Annex III assessment. Be honest: a tool that effectively ranks candidates is not "preparatory".

9. Have you documented and registered an exemption?

If you rely on Article 6(3), Article 6(4) requires you to document the assessment before placing the system on the market or putting it into service, register it under Article 49(2) and hand the documentation to authorities on request.

10. Who is the provider?

Classification follows the provider. If you put your name on a high-risk system, substantially modify one, or repurpose a general-purpose AI system for an Annex III use, Article 25 treats you as the provider. Our guide to AI provider vs deployer covers these switches.

What if the answer is "high-risk"?

Start building compliance into the product now rather than retrofitting it in late 2027. The first harmonised standard, EN 18286 on quality management, was cited in the Official Journal by Implementing Decision (EU) 2026/2274 on 14 October 2026, and SMEs can implement parts of the quality management system in a simplified way. Our step-by-step guide to AI Act compliance by design shows how to sequence the work. High-risk systems that process personal data will almost always need a GDPR impact assessment too; see our guide to a DPIA for AI features.

How reliable is this classification today?

Reasonably, but not final. Article 6 asked the Commission for guidelines with practical examples by 2 February 2026. The Commission published draft guidelines on 19 May 2026 and ran a consultation; the final text was still pending at the time of writing. Keep your written reasoning and revisit it once the guidelines are adopted.

Frequently asked questions

Is a chatbot high-risk under the EU AI Act?

Usually not. A general customer-service chatbot falls under the transparency rules in Article 50, not the high-risk regime. It becomes high-risk only if it is used for an Annex III purpose, such as deciding on job applicants or assessing eligibility for credit or public benefits.

Is AI that screens CVs always high-risk?

AI used to screen or filter applications, analyse candidates or place targeted job adverts is listed in Annex III point 4. It could only fall outside high-risk status under Article 6(3), for example a narrow procedural task, and never if it profiles candidates. Most CV-ranking tools profile applicants, so assume high-risk unless you can document otherwise.

Is fraud detection high-risk?

Annex III lists AI used to evaluate creditworthiness or establish credit scores, but expressly excludes AI used to detect financial fraud. A fraud model that does not also decide creditworthiness is therefore not high-risk on that ground.

Do I need to register a system I have decided is not high-risk?

Yes, if it is an Annex III system you are exempting under Article 6(3). Article 6(4) requires you to document the assessment before placing the system on the market and to meet the registration obligation in Article 49(2). Authorities can ask for your documentation.

When will the Commission's final high-risk guidelines be published?

The Commission published draft guidelines on classifying high-risk AI systems on 19 May 2026 and consulted on them. A final version had not been published at the time of writing, so check the Commission's AI Act pages before relying on any example.

If you would like a second opinion on how your AI product classifies before you commit to a roadmap, book a free 30-minute call.

Cover photo by RDNE Stock project on Pexels.