Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act)
Source URL: https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=OJ:L_202401689
Publication reference: OJ L 2024/1689, 12.7.2024
Date of adoption: 13 June 2024
Date of entry into force: 1 August 2024
CHAPTER I — GENERAL PROVISIONS
Article 1 — Subject matter
This Regulation lays down harmonised rules on artificial intelligence (AI) with the following objectives: (a) to ensure that AI systems placed on the market, put into service or used in the Union are safe and respect fundamental rights and Union values; (b) to ensure a high level of protection of health, safety and fundamental rights of natural persons against any harm caused by the design, development and use of AI systems; (c) to lay down a single set of requirements applicable to AI systems placed on the market, put into service or used in the Union, and to provide for harmonised conditions for the development, placement on the market, putting into service or use of AI systems in the Union; (d) to ensure the proper functioning of the internal market and to support innovation; (e) to facilitate the development of a single market for trustworthy AI applications and to promote the uptake of human-centric and trustworthy artificial intelligence, while ensuring a high level of protection of health, safety, fundamental rights enshrined in the Charter of Fundamental Rights of the European Union (“the Charter”) and the principles of non-discrimination, data protection, privacy, security and intellectual property rights; (f) to promote the development and use of AI in conformity with the Charter values, including fundamental rights, and the principles of non-discrimination, fairness, transparency, accountability and reliability; (g) to protect fundamental rights and freedoms of natural persons, in particular their right to dignity, non-discrimination, data protection, privacy and the right to property, against any harm caused by AI systems; (h) to ensure that AI systems are designed and used in a manner that ensures transparency, accountability, non-discrimination and respect for privacy.
Article 2 — Scope
1. This Regulation applies to: (a) providers placing on the market or putting into service AI systems, irrespective of whether those providers are established within the Union or in a third country; (b) deployers of AI systems that have their place of establishment or are located within the Union; (c) providers and deployers of AI systems that are based in a third country, where the output produced by the AI system is used in the Union.
2. This Regulation does not apply to AI systems: (i) developed or used exclusively for military, defence or national security purposes; (ii) developed or used exclusively for scientific research and development purposes; (iii) released under free and open-source licences, unless they are placed on the market or put into service as high-risk AI systems or used in safety-critical applications; (iv) intended to be used by public authorities in the territory of a Member State for activities outside the scope of this Regulation.
Article 3 — Definitions
For the purposes of this Regulation, the following definitions apply:
(1) ‘AI system’ means a machine-based system that is designed to operate with varying levels of autonomy and that can, for explicit or implicit objectives, generate outputs such as predictions, recommendations, or decisions influencing physical or virtual environments.
(2) ‘risk’ means the combination of the probability of an occurrence of harm and the severity of that harm.
(3) ‘provider’ means a natural or legal person, public authority, agency or other body that develops an AI system or that has an AI system developed and places that system on the market or puts it into service under its own name or trademark.
(4) ‘deployer’ means a natural or legal person, public authority, agency or other body using an AI system under its authority in a professional capacity.
(5) ‘importer’ means a natural or legal person established in the Union that places on the market or puts into service an AI system that bears the name or trademark of a natural or legal person established outside the Union.
(6) ‘distributor’ means a natural or legal person in the supply chain, other than the provider or the importer, that makes an AI system available on the Union market.
(7) ‘operator’ means a provider, product manufacturer, deployer, importer, distributor or authorised representative.
(8) ‘conformity assessment’ means the process of verifying whether the requirements set out in this Regulation relating to an AI system have been fulfilled.
(9) ‘notified body’ means a conformity assessment body designated in accordance with this Regulation.
(10) ‘substantial modification’ means a change to the AI system after its placing on the market or putting into service which is not foreseen or planned by the provider.
(11) ‘placing on the market’ means the first making available of an AI system on the Union market.
(12) ‘putting into service’ means the first use of an AI system within the Union by the provider or the deployer for its intended purpose.
(13) ‘intended purpose’ means the use for which an AI system is intended by the provider, including the specific application, conditions and purposes for which the AI system has been designed and developed.
(14) ‘reasonably foreseeable misuse’ means the use of an AI system in a way that is not in accordance with its intended purpose, but which may result from reasonably foreseeable human behaviour or interaction with the AI system.
(15) ‘safety component’ means a part or sub-element of a product or an AI system that fulfils a safety function.
(16) ‘biometric data’ means personal data resulting from specific technical processing relating to the physical, physiological or behavioural characteristics of a natural person.
(17) ‘biometric identification’ means the automated recognition of physical, physiological, behavioural or psychological features of an individual.
(18) ‘biometric verification’ means the automated verification of the identity of natural persons by comparing biometric data with previously provided biometric data.
(19) ‘real-time biometric identification’ means the use of biometric identification where the biometric data is collected and processed at the time of the identification.
(20) ‘post biometric identification’ means the use of biometric identification where the biometric data is collected prior to the identification.
(21) ’emotion recognition system’ means an AI system for the purpose of identifying or inferring emotions, feelings, thoughts, intentions or behavioural patterns of individuals on the basis of their biometric data.
(22) ‘biometric categorisation’ means the categorisation of natural persons into specific categories according to their biometric data.
(23) ‘remote biometric identification’ means the use of biometric identification where the biometric data is collected remotely and processed without the physical presence of the individual.
(24) ‘critical infrastructure’ means the assets, systems, facilities, networks, processes or other elements that are essential for the maintenance of vital societal functions, health, safety, security, economic or social well-being of people.
(25) ‘general-purpose AI model’ means an AI model that is trained on broad data at scale, is designed for generality of output, and can be adapted to a wide range of distinctive tasks.
(26) ‘systemic risk’ means a risk that is specific to the high-impact capabilities of general-purpose AI models, having a significant impact on the Union market due to its reach, severity, or other characteristics.
(27) ‘high-impact capabilities’ means capabilities that are sufficient to produce a significant impact on the Union market.
(28) ‘floating point operations per second (FLOPs)’ means a measure of computational performance used to quantify the resources required to train an AI model.
CHAPTER II — PROHIBITED AI PRACTICES
Article 5 — Prohibited AI Practices
1. The following AI practices shall be prohibited:
(a) the placing on the market, the putting into service or the use of an AI system that deploys subliminal techniques beyond a person’s consciousness or purposefully manipulative or deceptive techniques, with the objective or the effect of materially distorting the behaviour of an individual or a group of individuals by appreciably impairing their ability to make an informed decision, thereby causing them to take a decision they would not have otherwise taken, in a manner that causes or is likely to cause them harm;
(b) the placing on the market, the putting into service or the use of an AI system that exploits any of the vulnerabilities of natural persons or specific groups due to their age, disability or specific social or economic situation, with the objective or the effect of materially distorting the behaviour of that person or group of persons, in a manner that causes or is likely to cause that person or group of persons harm;
(c) the placing on the market, the putting into service or the use of AI systems for social scoring of natural persons, i.e., the evaluation or classification of the trustworthiness of natural persons based on their social behaviour or personal or personality characteristics, with the resulting detrimental or unfavourable treatment of persons in contexts unrelated to the contexts in which the data was originally generated or collected, or in an unjustified manner disproportionate to the context;
(d) the placing on the market, the putting into service or the use of AI systems for the purpose of assessing the risk of an individual committing an offence, solely based on profiling or personality and behavioural characteristics, without any assessment of the actual involvement of the individual in a criminal activity based on objective, verifiable facts and without it being linked to a serious criminal offence;
(e) the placing on the market, the putting into service or the use of AI systems that expand facial image databases, or create new facial image databases, through untargeted scraping of facial images from the internet or CCTV footage;
(f) the placing on the market, the putting into service or the use of AI systems to infer emotions of a natural person in the areas of workplace and education institutions, except where the use of the AI system is intended by the deployer to be used for medical or safety reasons;
(g) the placing on the market, the putting into service or the use of AI systems for biometric categorisation that categorises natural persons individually or jointly based on their biometric data to deduce or infer their race, political opinions, trade union membership, religious or philosophical beliefs, sex life or sexual orientation;
(h) the use of ‘real-time’ remote biometric identification systems in publicly accessible spaces for the purposes of law enforcement, unless and to the extent that such use is strictly necessary for one of the following objectives: (i) the targeted search for a missing person; (ii) the prevention of a substantial and imminent threat to the life of a natural person or a serious and imminent terrorist attack; (iii) the localisation or identification of a person suspected of having committed a serious criminal offence, with a maximum duration of the identification operation of 24 hours; or (iv) the prevention, investigation, detection or prosecution of a serious criminal offence or the execution of a criminal penalty, including the safeguarding against and the prevention of serious threats to public security.
2. The use of AI systems referred to in paragraph 1, point (h), in publicly accessible spaces for law enforcement purposes shall be authorised by a judicial or independent administrative authority of the Member State where the use takes place.
CHAPTER III — HIGH-RISK AI SYSTEMS
Article 6 — Classification rules for high-risk AI systems
1. An AI system shall be considered to be high-risk where one of the following conditions is met:
(a) the AI system is intended to be used as a safety component, or as a product, or the AI system is itself a product, covered by the Union harmonisation legislation listed in Annex I; or
(b) the AI system is intended to be used in any of the following areas and the AI system poses a significant risk of harm to the health, safety or fundamental rights of natural persons:
(i) biometric identification, categorisation and detection, except for the verification of biometric identity;
(ii) management and operation of critical infrastructure;
(iii) education and vocational training;
(iv) employment, workers management and self-employment;
(v) access to and enjoyment of essential private services and public services and benefits;
(vi) law enforcement;
(vii) migration, asylum and border control management;
(viii) administration of justice and democratic processes.
2. AI systems referred to in paragraph 1 shall be considered high-risk if they pose a significant risk of harm to the health, safety or fundamental rights of natural persons, including by materially influencing the decision-making resulting in the provision of essential private or public services or benefits to natural persons or groups of persons.
CHAPTER IV — GENERAL AI MODEL OBLIGATIONS
Article 50 — Classification Rules for General-Purpose AI Models
1. A general-purpose AI model shall be presumed to have systemic risk if it has high-impact capabilities, as determined by the Commission in accordance with paragraph 2.
2. A general-purpose AI model shall be considered to have high-impact capabilities if:
(a) it has the capability to produce output that matches or exceeds the performance of the most advanced models across a wide range of tasks;
(b) it is trained using a total amount of compute greater than 10^25 floating point operations (FLOPs).
3. The Commission may adopt delegated acts in accordance with Article 97 to amend the thresholds in this Article, or to add new thresholds, based on the latest technological developments, such as algorithmic improvements or increased hardware efficiency.
Article 51 — Obligations for Providers of General-Purpose AI Models
1. Providers of general-purpose AI models shall:
(a) draw up and keep up-to-date the technical documentation of the model, including its training process and evaluation results;
(b) provide information and documentation to the AI Office and national competent authorities on the model, its capabilities, and the results of evaluations and testing;
(c) implement a policy to ensure compliance with Union copyright law, in particular Article 4(3) of Directive (EU) 2019/790;
(d) establish a policy to ensure the protection of personal data, in particular the processing of personal data used in the development and testing of the AI model.
Article 52 — Obligations for Providers of General-Purpose AI Models with Systemic Risk
1. In addition to the obligations set out in Article 51, providers of general-purpose AI models with systemic risk shall:
(a) perform state-of-the-art model evaluations, including evaluations conducted in accordance with standardised protocols and benchmarks, to assess and mitigate systemic risks at the level of the Union;
(b) assess and mitigate possible systemic risks at Union level, including their sources, including but not limited to the development of the model, the training data, the results of model evaluations, and other relevant factors;
(c) track and report serious incidents and possible corrective measures to the AI Office and, as appropriate, to national competent authorities;
(d) ensure an adequate level of cybersecurity, physical security, and access controls to protect the model weights, infrastructure, and other relevant assets.
CHAPTER IX — PENALTIES
Article 99 — Penalties
1. Member States shall lay down the rules on penalties applicable to infringements of this Regulation and shall take all measures necessary to ensure that they are implemented. The penalties provided for shall be effective, proportionate and dissuasive. They shall take into account the nature, gravity and duration of the infringement, the size of the undertaking, the nature of the personal data or other data affected by the use, the extent of the damage, including the financial damage, suffered by natural persons or legal persons, the intentional or negligent character of the infringement and the degree of responsibility of the natural or legal person responsible for the infringement.
2. The maximum administrative fines for infringements of this Regulation shall be:
(a) EUR 35 000 000 or, if the offender is an undertaking, up to 7% of its total worldwide annual turnover of the preceding financial year, whichever is higher, for infringements of Article 5;
(b) EUR 15 000 000 or, if the offender is an undertaking, up to 3% of its total worldwide annual turnover of the preceding financial year, whichever is higher, for infringements of other provisions of this Regulation;
(c) EUR 7 500 000 or, if the offender is an undertaking, up to 1% of its total worldwide annual turnover of the preceding financial year, whichever is higher, for infringements of the following provisions: (i) Article 5(2); (ii) Articles 51 and 52; (iii) Article 59; (iv) Article 60; (v) Article 61; (vi) Article 62; (vii) Article 63; (viii) Article 64; (ix) Article 65; (x) Article 67; (xi) Article 68; (xii) Article 69; (xiii) Article 70; (xiv) Article 71; (xv) Article 72; (xvi) Article 73; (xvii) Article 74; (xviii) Article 75.
3. When deciding whether to impose an administrative fine and when deciding on the amount of the administrative fine, due regard shall be given to the following:
(a) the nature, gravity and duration of the infringement;
(b) the size of the undertaking, in particular the number of employees and the total worldwide annual turnover of the preceding financial year;
(c) the nature of the personal data or other data affected by the use of the AI system;
(d) the extent of the damage, including the financial damage, suffered by natural persons or legal persons;
(e) the intentional or negligent character of the infringement;
(f) the degree of responsibility of the natural or legal person responsible for the infringement;
(g) any previous infringements by the natural or legal person responsible for the infringement;
(h) the degree of cooperation with the competent authorities;
(i) the manner in which the infringement became known to the competent authorities, in particular whether, and if so to what extent, the natural or legal person responsible for the infringement notified the infringement to the competent authorities;
(j) any other aggravating or mitigating factor applicable to the circumstances of the case.
4. Member States shall ensure that administrative fines are imposed in accordance with this Article.
5. The imposition of criminal penalties for infringements of this Regulation and the relevant national criminal laws of Member States shall not affect the imposition of administrative fines under this Regulation.
Done at Brussels, 13 June 2024.
For the European Parliament
The President
R. METSOLA
For the Council
The President
M. MICHEL
