EU AI Act Article 6 (High-Risk Classification Rules)

Article 6 of the EU AI Act (Regulation (EU) 2024/1689) is the classification test that determines whether an AI system is legally “high-risk,” triggering Title III obligations under Articles 9-15 — risk management, data governance, technical documentation, logging, transparency, and human oversight. Article 6 itself imposes no obligations; it only decides whether they apply.

Two routes to high-risk classification

Route 1: Annex I product-safety route (Article 6(1))

An AI system is high-risk if it is intended to be used as a safety component of a product, or is itself a product, covered by the EU harmonisation legislation listed in Annex I (for example, medical devices, machinery, toys, lifts, and other CE-marked product categories) and that product is required to undergo third-party conformity assessment before it can be placed on the market under that sectoral legislation. This route ties AI classification to pre-existing EU product-safety law rather than creating a new standalone test.

Route 2: Annex III use-case route (Article 6(2))

AI systems falling into one of the eight functional categories listed in Annex III are automatically considered high-risk. The categories most relevant to research and higher-education institutions are:

• Education and vocational training — systems used to determine access or admission to an institution, evaluate learning outcomes, assess the appropriate level of education for a person, or monitor/detect prohibited behaviour during tests.
• Employment and worker management — recruitment and candidate-selection systems, and systems affecting the terms of a work relationship, promotion, termination, task allocation, or performance monitoring.
• Essential services — eligibility for public benefits or healthcare, creditworthiness evaluation, life/health insurance risk pricing, and emergency-call classification/dispatch.

The remaining Annex III categories — biometric identification, critical infrastructure, law enforcement, migration/asylum/border control, and administration of justice/democratic processes — are less commonly relevant to a typical research administration context but apply directly to institutions running biometric access-control research, security-related research, or work with law-enforcement/border-agency partners.

The narrow exception (Article 6(3)) and its override

An Annex III system is not high-risk if it does not pose a significant risk of harm to health, safety, or fundamental rights, and it does at least one of the following: performs a narrow procedural task; improves the result of a previously completed human activity; detects decision-making patterns or deviations from prior patterns without replacing or influencing human assessment without proper review; or performs a preparatory task for an assessment relevant to an Annex III use case. This exception is deliberately narrow — a screening tool that materially shapes an admissions or hiring outcome generally will not qualify. Critically, Article 6(3) contains a hard override: a system that performs profiling of natural persons is always considered high-risk regardless of whether it otherwise meets one of those four conditions.

A provider relying on this exception must document that assessment before the system is placed on the market or put into service, and register it under Article 49(2) (Article 6(4)). This is a real compliance step, not an informal judgment call — the documentation obligation applies even when the conclusion is “not high-risk.”

Timeline — verify before relying on any date

The Act’s original schedule set Annex III (use-case) high-risk obligations to apply from 2 August 2026 and Annex I (product-embedded) high-risk obligations from 2 August 2027. A “Digital Omnibus on AI” simplification package reached political agreement between the European Parliament and Council on 7 May 2026 (Parliament endorsement 16 June 2026, Council approval 29 June 2026) that would defer the Annex III compliance date to 2 December 2027 and the Annex I compliance date to 2 August 2028. As of this writing, formal publication of the amending act in the Official Journal of the EU — the step that makes revised dates legally binding — was still pending confirmation. Given how fast this is moving, verify the current status directly against the European Commission’s own AI Act implementation timeline (ai-act-service-desk.ec.europa.eu) before treating either date as settled for any compliance-critical decision.

Separately, the Commission is required under Article 6(5) to issue guidelines with practical examples of high-risk and non-high-risk use cases; draft classification guidelines went through public consultation (closing 23 June 2026) and were expected to finalize in the following weeks as of this writing — check for the published final version before relying on a specific worked example from the draft.

How this interacts with the research exemption

Article 6 classification is a separate question from the Act’s Article 2(6)/2(8) exemption for AI systems developed and used solely for scientific research and development prior to being placed on the market or put into service. A tool can sit entirely inside a high-risk Annex III category (e.g. an admissions-screening model) and still fall outside the Act’s substantive obligations for as long as it stays inside that research exemption — the obligations attach only once, and if, the system is placed on the market or put into service for real institutional use. See CASRAI’s guide to EU AI Act obligations and exemptions for research organizations for how that boundary is drawn in practice, and the EU AI Act Article 10 (Data and Data Governance) entry for one of the substantive obligations that attaches once a system does clear the Article 6 threshold.

Frequently asked questions

Does Article 6 itself require anything of an AI provider?

Not directly — Article 6 is a classification test. The substantive obligations (risk management, data governance, technical documentation, human oversight, and so on) live in the other provisions of Title III, Chapter 2, and only apply once a system is classified as high-risk under Article 6.

Is a university admissions AI tool automatically high-risk?

If it is intended to determine access or admission to an educational institution and is placed into real institutional use, it falls within the Annex III education category and is high-risk under Article 6(2), unless the narrow Article 6(3) exception applies and is documented — which is unlikely for a tool that materially influences an admissions decision, and never available if the tool profiles applicants.

What is the difference between the Annex I and Annex III routes?

Annex I ties classification to whether a product the AI is embedded in already requires third-party conformity assessment under separate EU product-safety law (e.g. medical devices). Annex III is a standalone list of high-risk use cases that applies regardless of whether any other EU product law is in play.