Skip to content

EU AI Act — Commission guidance

The European Commission has issued interpretive guidance on three areas of the EU AI Act: the definition of an AI system (Article 3(1)), the prohibited AI practices (Article 5), and — in draft form — the classification of high-risk AI systems (Article 6). This page is the entry point for the docs.modulos.ai treatment of those five source documents.

Status

Commission interpretive guidance, not binding law. The EU AI Act text and any Court of Justice of the European Union (CJEU) interpretation prevail in case of conflict. The high-risk classification guidance on this site is currently a draft for stakeholder consultation closing 23 June 2026; its text may change before formal adoption.

Quick decision

TL;DR

  • Commission guidance under Article 96 AI Act is soft law — interpretive, not binding. The Regulation text and the CJEU prevail.
  • Three sets, five Commission source documents: the AI-system definition (C(2025) 5053, final, 29 July 2025); prohibited practices (C(2025) 5052, final, 29 July 2025); and the draft high-risk classification guidelines (May 2026, three companion documents, consultation closes 23 June 2026).
  • The two final Communications are stable references. The draft high-risk guidance may change before adoption — flag this in any compliance assessment that relies on it.
  • Market surveillance authorities will read the Commission's interpretation closely; documenting how your own assessment relates to it is good practice.
  • This site keeps the Regulation text on the primary EU AI Act spokes and the Commission's reading on the four commission-guidance/ pages. The split is intentional.

What Commission guidance is and is not

Article 96(1) AI Act mandates the Commission to develop guidelines on the application of the Regulation. These guidelines have a specific legal character:

  • They are adopted by the Commission as Communications, not as legislative acts. They go through Commission internal procedures and stakeholder consultation, not the ordinary legislative procedure.
  • They state the Commission's interpretation. They do not amend the Regulation. They cannot create new obligations beyond the Regulation text, and they cannot remove obligations the Regulation imposes.
  • They are not binding on Member States, national market surveillance authorities, providers, deployers, courts, or the CJEU. The CJEU has the final word on the meaning of EU law.
  • In practice, they shape how market surveillance authorities approach enforcement and how courts read ambiguous Regulation text. A defensible compliance position usually engages with the Commission's reading even where it ultimately departs from it.

The Commission itself states this status explicitly in each Communication — see for example the AI-system definition guidance at paragraph (7): "The Guidelines are not binding. Any authoritative interpretation of the AI Act may ultimately only be given by the Court of Justice of the European Union (CJEU)."

The guidance source documents

DocumentIdentifierStatusDateSource
Commission Guidelines on the definition of an AI systemC(2025) 5053 finalFinal Communication29 July 2025AI Act Service Desk PDF
Commission Guidelines on prohibited AI practicesC(2025) 5052 finalFinal Communication29 July 2025AI Act Service Desk PDF
Draft Commission Guidelines on the classification of high-risk AI systems — general principles(consultation draft)Draft for stakeholder consultationMay 2026, consultation closes 23 June 2026Commission newsroom PDF
Draft Commission Guidelines on the classification of high-risk AI systems — Annex I route(consultation draft)Draft for stakeholder consultationMay 2026, consultation closes 23 June 2026Annex I PDF
Draft Commission Guidelines on the classification of high-risk AI systems — Annex III route(consultation draft)Draft for stakeholder consultationMay 2026, consultation closes 23 June 2026Annex III PDF

The three draft high-risk documents are companions and are intended to be read together. This site keeps the classification framework on one page and the worked examples on a separate page because the framework is decision-logic-dense and the examples are reference material.

Primary source

Commission Communications adopted on 29 July 2025 for the definition and prohibited-practices guidelines (C(2025) 5053 and C(2025) 5052). Draft high-risk classification guidelines published on 19 May 2026 for stakeholder consultation closing 23 June 2026. The final Communications and the three draft high-risk consultation documents © European Union. PDFs available via the AI Act Service Desk and via the Commission newsroom redirection links referenced in the table above.

How to use the guidance in a Modulos compliance programme

Each guidance document anchors to a specific Modulos requirement on the EU AI Act framework template:

Guidance documentArticleModulos requirementCode
AI-system definition (final)Article 3(1)AI System Classification (app)MRF-38
Prohibited AI practices (final)Article 5Prohibited AI Practices (app)MRF-119
High-risk classification (draft) — frameworkArticle 6(1), 6(2), 6(3)AI System Classification (app); AI System Classification Exemption (app)MRF-38, MRF-111
High-risk classification (draft) — worked examplesArticle 6 + Annex I/IIIAI System Classification (app); AI System Classification Exemption (app)MRF-38, MRF-111

The Modulos framework templates are OFF-1 (organisation) and MFF-1 (AI application). For details see Operationalizing the EU AI Act in Modulos.

Cross-framework mapping (preview)

Commission guidanceAdjacent reading
Definition (Article 3(1))OECD updated AI-system definition (2024); ISO/IEC 22989 vocabulary; NIST AI RMF Section 2 (Characteristics of trustworthy AI).
Prohibited practices (Article 5)UNESCO Recommendation on the Ethics of AI; OECD AI Principles (2019/2024); Council of Europe Framework Convention on AI.
High-risk classification (Article 6)OECD high-risk classification framework; NIST AI RMF MAP function (system characterisation); national risk-based AI laws (e.g. Colorado SB 24-205, Canada AIDA draft).

Source attribution

Commission Communication C(2025) 5052 final, 29 July 2025 — Commission Guidelines on prohibited artificial intelligence practices established by Regulation (EU) 2024/1689 (AI Act). Commission Communication C(2025) 5053 final, 29 July 2025 — Commission Guidelines on the definition of an artificial intelligence system established by Regulation (EU) 2024/1689 (AI Act). Draft Commission Guidelines on the classification of high-risk artificial intelligence systems under Article 6 of the EU AI Act for stakeholder consultation, May 2026 (three companion documents). The final Communications and draft high-risk consultation documents © European Union. PDFs available via the AI Act Service Desk and the Commission newsroom.

Disclaimer

This page is for general informational purposes and does not constitute legal advice.