EU AI Act
From legal text to evidence

1. 01

   The article

   The Official Journal text, transcribed word for word and checked against a hash of the source it came from. Nothing is paraphrased, and nothing is written from memory.

2. 02

   The clause

   Each article becomes one clause in the catalogue, carrying its official heading, its citation and any deadline the text sets. This is the layer the platform reads, and it is shared across every tenant.

3. 03

   Your obligation

   You adopt the clauses that apply to you, and each becomes a tracked obligation of your own: an owner, an applicability decision, a review date, and the evidence that it is being met.

Where an article sets a reporting clock, the deadline the Incident Centre counts down is computed from that clause and starts when the text says it starts, rather than from a duration typed into the product.

1. 13 Jun 2024

   Adopted by the Parliament and the Council

2. 12 Jul 2024

   Published in the Official Journal

3. 1 Aug 2024

   Entered into force

   Art. 113

4. 2 Aug 2026

   General application begins

   Art. 113

The regulation runs to 113 articles. Pritect decomposes 16 of them clause by clause into obligations you can adopt, own and evidence: the requirements that attach to a high-risk AI system, the duties that fall on the provider that builds one and on the organisation that deploys it, and the transparency duties that apply whether or not a system is high risk. The rest are deliberately not decomposed, because they are not duties you discharge. Articles 1 to 8 set the subject matter, scope, definitions, prohibited practices and the high-risk classification rules, Articles 22 to 24 address authorised representatives, importers and distributors, Articles 28 to 49 govern notifying authorities, notified bodies, standards and conformity assessment, Articles 51 to 56 cover general-purpose AI models, Articles 57 to 72 cover innovation measures, governance, the EU database and post-market monitoring, and Articles 74 to 113 govern market surveillance, remedies, penalties and the final provisions. Article 15 and Article 73 are catalogued at clause level but sit outside the article list the AI governance review confirmed for this hub, so they are not listed here. The set grows as the catalogue is extended.

Article headings are reproduced verbatim from the Official Journal and stay in English in every language, because a citation has to remain quotable. They are transcribed from the published text; the subject-matter review that makes them authoritative inside the product is still pending.

Which AI systems count as high risk?

Article 6 sets the classification rules. A system is high risk where it is a safety component of a product covered by other Union law, or where it falls in one of the areas listed in Annex III, such as employment, education, essential services, law enforcement and migration. Pritect records the classification and the reasoning behind it against each system.

Are we a provider or a deployer?

The Act puts different duties on each. A provider develops a system, or has one developed, and places it on the market under its own name. A deployer uses a system under its own authority. Putting your own name on a system you bought can make you a provider under Article 25, which is why Pritect asks the question per system rather than per organisation.

Do we need a fundamental rights impact assessment?

Article 27 requires one from deployers that are public bodies or that provide public services, and from deployers of certain creditworthiness and insurance pricing systems, before a high-risk system is put into use. Pritect runs the screening and keeps the assessment attached to the system it belongs to.

When do the obligations start to apply?

Article 113 staggers commencement. The general date of application is the one shown in the timeline above, with the prohibited practices and the general provisions applying earlier, and the rules for high-risk systems that are safety components of regulated products applying later. Pritect tracks which date governs each system you have recorded.

What has to be disclosed to people?

Article 50 requires that people are told when they are interacting with an AI system, that synthetic audio, image, video and text is marked as artificially generated, and that emotion recognition and biometric categorisation are disclosed to the people exposed to them. These duties apply whether or not the system is high risk.

This page summarises publicly available legal text so you can orient yourself. It is not legal advice.