“It shall apply from 2 August 2026.” This sentence forms the second paragraph of Article 113 of Regulation (EU) 2024/1689 on artificial intelligence. It is followed by the word “However” and by four points, (a) to (d), which bring this date forward or push it back depending on the chapter. Two of these points were rewritten, and the last one added, by Regulation (EU) 2026/1744 of 8 July 2026.
AI Act obligations for businesses and their timetable are determined in three readings, carried out for each system: the role held (provider or deployer), the risk level, then the date that Article 113 attaches to the chapter concerned. The prohibitions, apart from two additions in 2026, and the AI literacy obligation have applied since 2 February 2025 (third paragraph, point (a)), the transparency obligations since 2 August 2026 (second paragraph), and the rules on high-risk systems apply from 2 December 2027 or 2 August 2028 (point (c)). Quotations come from the consolidated version of 27 July 2026; our entry on the AI Act defines the notions.
The same business can be a deployer here and a provider there
According to Article 3, the provider (point 3) develops an AI system or has it developed and places it on the market or puts it into service under its own name. The deployer (point 4) is the person or body “using an AI system under its authority except where the AI system is used in the course of a personal non-professional activity”.
A business is the deployer of the case-processing support software it buys from a software vendor. In our reading, it is the provider of the pricing engine it has developed and puts into service under its own name: point 11 of the same article includes in putting into service the supply of a system for first use “for own use”. The Commission’s guidelines on Article 50 (C/2026/6005, point 11), which are not binding, give this example for an interactive AI system.
Article 25(1) makes a deployer the provider of a high-risk system, bound by Article 16, in particular when it modifies the intended purpose of a system in such a way that the system becomes high-risk. The classification is a matter for the business and its legal advisers.
AI Act obligations for businesses by risk level
Article 4 does not depend on any classification. In its wording resulting from Regulation (EU) 2026/1744, it asks providers and deployers to take measures to foster the development of AI literacy among their staff. It adds: “This obligation does not require providers or deployers to ensure a specific level of AI literacy by an individual.”
Article 5 prohibits practices. Depending on the point, it covers placing on the market, putting into service and use: the deployer is covered as well as the provider.
For a high-risk system (Article 6, Annexes I and III), the provider bears the obligations of Article 16, including compliance with the requirements of Section 2, conformity assessment and CE marking. The deployer bears those of Article 26, including use in accordance with the instructions for use and human oversight. It keeps the logs under its control for a period “of at least six months”, unless otherwise provided in applicable law (paragraph 6). Article 27 requires a fundamental rights impact assessment from certain deployers, including those of a creditworthiness assessment system (Annex III, point 5(b)).
Article 50 allocates the transparency obligations. The provider ensures that people know they are interacting with an AI system and marks synthetic content in a machine-readable format. The deployer discloses deep fakes in particular.
Chapter V covers providers of general-purpose AI models, including large language models. Its Article 53(1)(b) requires them to make documentation available to providers of systems that intend to integrate their model.
The application timetable as read in Article 113
| Applicable from | Provisions | Reference in Article 113 |
|---|---|---|
| 2 February 2025 | Chapters I and II, including AI literacy (Article 4) and prohibited practices (Article 5) | Third paragraph, point (a) |
| 2 August 2025 | Chapter V (general-purpose AI models), Chapter XII (penalties) except Article 101, Chapter III Section 4, Chapter VII and Article 78 | Third paragraph, point (b) |
| 2 August 2026 | General date: the provisions not covered by any exception, including Chapter IV (Article 50) | Second paragraph |
| 2 December 2026 | Article 5(1), first subparagraph, points (ba) and (bb), and paragraphs 1a and 1b, added in 2026: they cover in particular realistic content depicting the intimate parts of an identifiable person without their consent | Third paragraph, point (a) |
| 2 December 2027 | Chapter III, Sections 1, 2 and 3, except Article 6(5), for the high-risk systems of Article 6(2) and Annex III | Third paragraph, point (c)(i) |
| 2 August 2028 | The same sections for the high-risk systems of Article 6(1) and Annex I | Third paragraph, point (c)(ii) |
Articles 16, 26 and 27, located in Section 3 of Chapter III, follow the last two rows. For Article 50, guidelines C/2026/6005 (point 153) also retain the general date. Point (d) makes Articles 102 to 110, which amend other Union acts, applicable from 27 July 2026.
Systems already in service also fall under Article 111
According to Article 111(2), and excluding the large-scale information systems of paragraph 1, the regulation applies to operators of high-risk systems placed on the market or put into service before the date of application of Chapter III “only if, as from that date, those systems are subject to significant changes in their designs”, without prejudice to Article 5. A specific deadline, 2 August 2030, applies to those intended to be used by public authorities.
Paragraph 4 gives providers of systems generating synthetic content placed on the market before 2 August 2026 until 2 December 2026 for Article 50(2).
The date a high-risk system was put into service and the history of its modifications should therefore be kept.
The consequences of this reading for a project
What follows is a matter of method: the regulation does not prescribe how a project is run.
At the framing stage, the project sheet records the business’s role, the regime under which the system falls and the date of application, with the point of Article 113 on which it rests. A change of purpose reopens them: the intended purpose drives the classification and, through Article 25, the role.
When buying a high-risk system, the instructions for use are requested before signature, since Article 26 binds the deployer to that document, and the contract specifies who holds the logs.
In the schedule, the go-live milestone is compared with the dates in point (c). The obligations of Articles 16 or 26 form part of the cost of the high-risk system, and therefore of the trade-off between use cases conducted by AI strategy.
Keeping these records for the whole estate requires a register and a validation circuit: our AI governance page describes its content, and our article on implementing AI governance describes how it unfolds.
The next deadline, at the time of writing, is 2 December 2026. It concerns the prohibitions added to Article 5 and, through Article 111(4), the marking of synthetic content by systems already on the market. A business that generates content with such a system can ask its provider now how the outputs are marked.