Skip to content
Technology transformation expertise firm · ARDIANE Group Innovate, secure, succeed

Glossary

AI Act: the European regulation on artificial intelligence

Domain

Artificial intelligence

In brief

The AI Act is Regulation (EU) 2024/1689 of 13 June 2024, which lays down harmonised rules for placing artificial intelligence systems on the market and using them in the European Union. It grades obligations according to the level of risk of the use and allocates them between several actors, first and foremost the provider and the deployer.

Page contents

Definition of the AI Act

The AI Act is the common name of Regulation (EU) 2024/1689 of 13 June 2024 laying down harmonised rules on artificial intelligence. In French, it is also called the règlement sur l’intelligence artificielle, abbreviated to RIA. It entered into force on 1 August 2024 and its application is staggered. Like any European regulation, it applies directly in the Member States, without any transposing law.

The text governs the placing on the market, putting into service and use of AI systems in the Union. It does not regulate a technology as such. It grades obligations according to the risk that the use poses to people’s health, safety and fundamental rights.

What the AI Act is for

The regulation pursues two aims: protecting people against dangerous uses of AI, and giving the internal market a single framework rather than legislation by Member State.

Its scope, defined in Article 2, extends beyond the borders of the Union. It covers providers that place a system on the European market, wherever they are established, deployers established in the Union, and actors in third countries where the output produced by the system is used in the Union. Several uses are excluded, including exclusively military, defence or national security purposes, and strictly personal, non-professional use.

The risk-based approach: four regimes

Prohibited practices (Article 5). Certain uses are banned, among them manipulative techniques that distort a person’s behaviour to their detriment, the exploitation of vulnerabilities related to age or disability, social scoring and emotion recognition in the workplace or in education, subject to the exceptions provided for in the text.

High-risk systems (Article 6). Two routes lead to this category. The first covers systems that are safety components of products (or are themselves products) covered by the Union harmonisation legislation listed in Annex I, where that product is subject to third-party conformity assessment. The second covers the uses listed in Annex III, in areas such as biometrics, critical infrastructure, education, employment, access to essential services, law enforcement, migration management or justice. These systems must meet detailed requirements: risk management, data governance, technical documentation, logging, information for deployers, human oversight, accuracy, robustness and cybersecurity.

Transparency obligations (Article 50). A person must know that they are interacting with an AI system. Artificially generated or manipulated content, including deepfakes, must be labelled as such.

General-purpose AI models (Chapter V). Their providers, including those of large language models, must in particular document the model and inform the actors who integrate it into their systems. Additional obligations apply to models presenting a systemic risk.

Systems that fall under none of these regimes are not subject to any specific obligation under this classification. Article 4, which requires providers and deployers to foster AI literacy among their staff, applies on the other hand whatever the level of risk.

Provider, deployer: who bears what

Article 3 defines the actors. The two main ones are as follows.

  • The provider develops an AI system or a general-purpose AI model, or has it developed, and places it on the market or puts it into service under its own name or trademark. For a high-risk system, it bears most of the compliance burden: conformity assessment, documentation, quality management system, post-market monitoring.
  • The deployer uses an AI system under its own authority, in a professional context. For a high-risk system, Article 26 requires it among other things to follow the instructions for use, to entrust human oversight to competent persons, to monitor the operation of the system and to keep the logs.

The regulation also recognises the importer, the distributor and the authorised representative. The boundary between roles is not fixed. Under Article 25, a deployer is treated as a provider if it markets a high-risk system under its own name, if it substantially modifies it, or if it changes its intended purpose to the point of bringing it into the high-risk category. A company that adapts a purchased tool must therefore check which role it is taking on.

AI Act and GDPR: not to be confused

The two texts do not have the same purpose. The GDPR, Regulation (EU) 2016/679, protects personal data, whatever the technology that processes it. The AI Act draws on product regulation: it concerns the AI system itself, whether or not it handles personal data.

They are cumulative. An AI system that processes personal data must comply with both, and the roles do not overlap: a deployer within the meaning of the AI Act is often a controller within the meaning of the GDPR, but this is not automatic.

Application timetable and penalties

The regulation applies progressively: the prohibitions apply first, followed by the rules on general-purpose AI models, then those on high-risk systems. The timetable is set out in Article 113, as amended by Regulation (EU) 2026/1744 of 8 July 2026. Chapters I and II, including the prohibitions in Article 5, have applied since 2 February 2025, with the exception of two prohibitions added by the 2026 regulation, which apply from 2 December 2026. The rules on general-purpose AI models have applied since 2 August 2025. The general date of application is 2 August 2026. By way of exception, the rules on high-risk systems apply from 2 December 2027 for the uses in Annex III and from 2 August 2028 for systems linked to the products in Annex I.

Penalties are provided for in Article 99, in the form of administrative fines whose ceiling varies according to the seriousness of the infringement. Supervision falls to the national market surveillance authorities and, for general-purpose AI models, to the AI Office set up within the European Commission. Since Regulation (EU) 2026/1744, Article 75 also gives the AI Office alone the supervision of certain AI systems, in particular those based on a general-purpose AI model developed by the same provider.

A concrete example: a tool for sorting job applications

Take a company that buys from a software vendor a program that ranks the applications it receives. Annex III lists among high-risk uses systems intended for recruitment, in particular for filtering applications and evaluating candidates. The tool therefore in principle falls into this category, subject to the exceptions in Article 6.

The vendor is the provider: the compliance of the system is its responsibility. The company is the deployer. It must follow the instructions, appoint people capable of checking and correcting the tool’s proposals, and inform candidates that such a system is involved in processing their application. If it has the tool extensively reconfigured or distributes it under its own brand, it may switch to the role of provider.

To organise the inventory and classification of your systems, see our page dedicated to AI governance.

FAQ

Frequently asked questions

Another question? Ask us

What is the AI Act?

The AI Act is the common name of Regulation (EU) 2024/1689 of 13 June 2024 laying down harmonised rules on artificial intelligence. In force since 1 August 2024, it governs the placing on the market, putting into service and use of AI systems in the Union, with obligations proportionate to the risk.

What levels of risk does the AI Act provide for?

The regulation distinguishes prohibited practices, high-risk AI systems subject to detailed requirements, and certain systems subject to transparency obligations. It also provides a specific regime for general-purpose AI models. Other systems are not subject to any specific obligation on this basis.

What is the difference between provider and deployer in the AI Act?

The provider 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 uses the system under its own authority in a professional context. The provider bears most of the compliance obligations, the deployer obligations of use and control.

Does the AI Act replace the GDPR?

No. The GDPR, Regulation (EU) 2016/679, protects personal data whatever the technology used. The AI Act governs AI systems, whether or not they process personal data. A single project may fall under both texts, whose obligations are cumulative.

A project to bring to completion?

A first conversation is enough to understand your challenge, identify the right people and tell you frankly what we can bring.