Compaia AI Act obligations

Sheets by article, obligation, annex and recital with roles, actions and expected documents.

Pages

Annexe I - Annex I — List of Union harmonization legislation Markdown
This list applies to all EU Member States as part of the Union's regulatory framework.
Article 1 — Subject matter Markdown
This regulation applies to all AI systems placed on the market or put into service within the European Union.
Recital (1) Markdown
European Union
Annexe II - Annex II — List of criminal offences referred to in Article 5(1), first subparagraph, point (h)(iii) Markdown
This list applies to all EU Member States as defined under the AI Act.
Article 2 - Article 2 — Scope Markdown
The AI Act defines its scope for EU and non-EU actors impacting the European market, with specific exclusions for military, national security, and research domains. It clarifies interactions with other European regulations.
Recital (2) Markdown
European Union
Annexe III - Annex III — High-risk AI systems referred to in Article 6(2) Markdown
Annex III details high-risk AI systems, including biometrics, critical infrastructure, education, employment, essential services, law enforcement, migration, and justice.
Article 3 - Article 3 — Definitions Markdown
Article 3 defines key terms of the AI Act, including AI systems, actor roles, and technical concepts. It also covers data, risk, and compliance procedures.
Recital (3) Markdown
The regulation applies to the European Union to ensure a uniform level of protection and prevent market fragmentation.
Annexe IV - Annex IV — Technical documentation referred to in Article 11(1) Markdown
Annex IV details the technical documentation required for AI systems, including their description, development, monitoring, and compliance with standards.
Article 4 - Article 4 — AI Literacy Markdown
This article applies to all providers and deployers of AI systems within the scope of the EU AI Act.
Recital (4) Markdown
The European Union
Annexe V - Annex V — EU Declaration of Conformity Markdown
This annex applies to all high-risk AI systems placed on the EU market.
Article 5 - Article 5 — Prohibited AI Practices Markdown
This article applies to all providers and deployers of AI systems within the European Union, regardless of their place of establishment, if the output produced by the system is used in the Union.
Recital (5) Markdown
This applies to AI systems operating within the European Union and those affecting individuals in the Union, regardless of where the provider is established.
Annexe VI - Annex VI — Conformity assessment procedure based on internal control Markdown
The conformity assessment procedure is based on internal control by the provider, including verification of the quality management system, examination of technical documentation, and consistency of design and monitoring processes.
Article 6 - Article 6 — Rules for the classification of AI systems as high-risk Markdown
This article applies to all AI systems placed on the EU market or put into service within the Union, as defined by the AI Act.
Recital (6) Markdown
European Union
Annexe VII - Annex VII — Conformity based on an assessment of the quality management system and an assessment of the technical documentation Markdown
This annex applies to providers of high-risk AI systems seeking conformity assessment under the EU AI Act.
Article 7 — Amendments to Annex III Markdown
Article 7 details the conditions for amending Annex III, adding or removing high-risk AI systems based on specific criteria.
Recital (7) Markdown
European Union
Annexe VIII - Annex VIII — Information to be submitted upon the registration of high-risk AI systems in accordance with Article 49 Markdown
Annex VIII details the information required for the registration of high-risk AI systems, categorized into three sections based on the relevant actors (providers and deployers).
Article 8 — Compliance with requirements Markdown
High-risk AI systems must comply with specific requirements, taking into account their intended purpose and the state of the art. Providers must ensure compliance with EU requirements.
Recital (8) Markdown
This recital applies to all EU Member States as part of the harmonized framework for the internal market.
Annexe IX - Annex IX — Information to be provided upon registration of high-risk AI systems listed in Annex III regarding testing in real-world conditions in accordance with Article 60 Markdown
This regulation applies to all EU Member States.
Article 9 - Article 9 — Risk Management System Markdown
The risk management system must be established, implemented, documented, and maintained for high-risk AI systems to ensure compliance with safety and fundamental rights requirements.
Recital (9) Markdown
This recital applies to all EU Member States as part of the EU AI Act regulatory framework.
Annexe X - Annex X — Union legislative acts concerning large-scale IT systems in the area of freedom, security and justice Markdown
Annex X details EU legislative acts concerning large-scale IT systems, including Schengen, visas, and Eurodac, for border management and police cooperation.
Article 10 - Article 10 — Data and data governance Markdown
Article 10 applies to providers of high-risk AI systems, requiring them to implement robust data governance practices throughout the lifecycle of the AI system.
Recital (10) Markdown
This recital applies to all EU Member States and entities subject to the AI Act.
Annexe XI - Annex XI — Technical documentation referred to in Article 53(1)(a) – technical documentation for providers of general-purpose AI models Markdown
This regulation applies to providers of general-purpose AI models operating within the scope of the EU AI Act.
Article 11 - Article 11 — Technical documentation Markdown
This article applies to all providers of high-risk AI systems within the scope of the EU AI Act.
Recital (11) Markdown
This Regulation should be without prejudice to the provisions on the liability of providers of intermediary services set out in Regulation (EU) 2022/2065 of the European Parliament and of the Council.
Annexe XII - Annex XII — Transparency information referred to in Article 53(1)(b) – technical documentation for providers of general-purpose AI models to downstream providers that integrate the model into their AI system Markdown
Annex XII details the transparency information required for general-purpose AI models, including their description, development, and usage.
Article 12 - Article 12 — Logging Markdown
This regulation applies to high-risk AI systems deployed within the European Union.
Recital (12) Markdown
This recital applies to all entities operating within the scope of the EU AI Act, providing the foundational definition of an AI system for regulatory compliance.
Annexe XIII - Annex XIII — Criteria for designating general-purpose AI models with systemic risk referred to in Article 51 Markdown
This annex applies to general-purpose AI models within the scope of the EU AI Act.
Article 13 - Article 13 — Transparency and provision of information to deployers Markdown
Article 13 mandates transparency for high-risk AI systems and requires a detailed set of instructions for deployers, covering characteristics, limitations, and control measures.
Recital (13) Markdown
This definition applies to all entities operating within the scope of the EU AI Act.
Article 14 - Article 14 — Human Oversight Markdown
Article 14 governs human oversight of high-risk AI systems to minimize risks, requiring proportionate measures and adapted interfaces.
Recital (14) Markdown
This recital applies to all EU Member States subject to the AI Act.
Article 15 - Article 15 — Accuracy, robustness and cybersecurity Markdown
This article applies to providers of high-risk AI systems within the scope of the EU AI Act.
Recital (15) Markdown
This definition applies to all EU member states under the scope of the AI Act.
Article 16 - Article 16 — Obligations of providers of high-risk AI systems Markdown
This article applies to all providers of high-risk AI systems operating within the European Union or placing such systems on the EU market.
Recital (16) Markdown
This regulation applies to the European Union and entities operating within or providing services to the EU market.
Article 17 — Quality Management System Markdown
This article applies to providers of high-risk AI systems operating within the European Union under the AI Act.
Recital (17) Markdown
This recital clarifies the functional definition of remote biometric identification systems, distinguishing between 'real-time' and 'post' systems, and excluding biometric verification used for authentication or secure access.
Article 18 — Record-keeping Markdown
Providers must maintain documentation for 10 years following the date on which the high-risk AI system is placed on the market.
Recital (18) Markdown
This recital clarifies the scope of emotion recognition systems within the AI Act, distinguishing between emotional states and physical or mental states.
Article 19 — Automatically generated logs Markdown
This regulation applies to all high-risk AI systems deployed within the European Union.
Recital (19) Markdown
The definition applies to any physical space accessible to an indeterminate number of natural persons, excluding online spaces and areas restricted to specific individuals by law or property owners, such as private offices, factories, or prisons.
Article 20 — Corrective measures and duty of information Markdown
This article applies to providers of high-risk AI systems operating within the scope of the EU AI Act.
Recital (20) Markdown
This recital applies to all providers, deployers, and relevant parties operating within the scope of the EU AI Act.
Article 21 — Cooperation with competent authorities Markdown
Providers of high-risk AI systems must cooperate with competent authorities upon request to demonstrate compliance with the AI Act.
Recital (21) Markdown
The regulation applies to providers of AI systems regardless of whether they are established in the Union or in a third country, and to deployers of AI systems established in the Union.
Article 22 — Authorized representatives of providers of high-risk AI systems Markdown
This article applies to providers of high-risk AI systems established outside the European Union who intend to place their systems on the EU market.
Recital (22) Markdown
The Regulation applies to AI systems even when not placed on the market in the Union, provided the provider or deployer is established in the Union, or the output of the AI system is used within the Union.
Article 23 — Obligations of importers Markdown
Article 23 details the obligations of importers of high-risk AI systems, including compliance checks, document retention, and cooperation with competent authorities.
Recital (23) Markdown
European Union
Article 24 — Obligations of distributors Markdown
This article applies to all distributors of high-risk AI systems operating within the European Union market.
Recital (24) Markdown
Where and to the extent that AI systems are placed on the market, put into service or used with or without modification for military, defence or national security purposes, these systems should be excluded from the scope of this Regulation.
Article 25 — Responsibilities along the AI value chain Markdown
This article applies to all providers and third parties involved in the AI value chain within the scope of the EU AI Act.
Recital (25) Markdown
The regulation excludes AI systems developed specifically for scientific research and development, as well as R&D activities conducted prior to market placement or service entry.
Article 26 - Article 26 — Obligations of deployers of high-risk AI systems Markdown
The AI Act imposes strict obligations on deployers of high-risk AI systems, including human oversight, monitoring, logging, stakeholder notification, and cooperation with competent authorities.
Recital (26) Markdown
The risk-based approach requires tailoring regulatory rules to the intensity and scope of risks posed by AI systems, including prohibitions for unacceptable practices, requirements for high-risk systems, and transparency obligations.
Article 27 - Article 27 — Fundamental rights impact assessment for high-risk AI systems Markdown
This requirement applies to deployers of high-risk AI systems as defined under the EU AI Act.
Recital (27) Markdown
EU
Article 28 — Notifying authorities Markdown
Member States are required to designate a notifying authority to oversee conformity assessment bodies, ensuring impartiality, technical competence, and the protection of fundamental rights.
Recital (28) Markdown
European Union
Article 29 — Application for notification of a conformity assessment body Markdown
The application must be submitted to the notifying authority of the Member State in which the conformity assessment body is established.
Recital (29) Markdown
This recital applies to all EU Member States as part of the AI Act framework.
Article 30 — Notification procedure Markdown
This article applies to all EU Member States responsible for notifying conformity assessment bodies under the AI Act.
Recital (30) Markdown
This regulation applies to all biometric categorization systems operating within the European Union that infer sensitive personal data.
Article 31 — Requirements for notified bodies Markdown
Notified bodies must be established under national law, possess legal personality, and meet strict criteria regarding organizational structure, independence, and technical competence to perform conformity assessments under the AI Act.
Recital (31) Markdown
The regulation prohibits AI systems used for social scoring to prevent discrimination and violations of fundamental rights.
Article 32 — Presumption of conformity with requirements for notified bodies Markdown
This article applies to all EU Member States as part of the AI Act framework.
Recital (32) Markdown
EU
Article 33 — Subsidiaries of notified bodies and subcontracting Markdown
Notified bodies must ensure that their subsidiaries and subcontractors comply with the requirements set out in Article 31 and must maintain full responsibility for all tasks performed by them.
Recital (33) Markdown
The regulation applies to the use of biometric identification systems by law enforcement authorities within the European Union, subject to specific, narrowly defined exceptions.
Article 34 — Operational obligations of notified bodies Markdown
This article applies to all notified bodies operating within the European Union under the AI Act framework.
Recital (34) Markdown
This recital applies to the use of remote biometric identification systems within the European Union under the AI Act.
Article 35 — Identification numbers and lists of notified bodies Markdown
The Commission assigns a unique identification number to each notified body and publishes an updated list of these bodies along with their activities.
Recital (35) Markdown
This recital applies to the use of real-time remote biometric identification systems by law enforcement in publicly accessible spaces within the European Union.
Article 36 — Changes to notifications Markdown
This article applies to all EU Member States and notified bodies operating under the AI Act.
Recital (36) Markdown
This applies to all EU Member States subject to the AI Act.
Article 37 — Challenging the competence of notified bodies Markdown
Article 37 governs the process for challenging the competence of notified bodies, including investigation procedures, confidentiality requirements, and corrective actions.
Recital (37) Markdown
The use of AI systems on the territory of a Member State is subject to the national law of that specific Member State, which may authorize, restrict, or prohibit such use.
Article 38 — Coordination of notified bodies Markdown
This article applies to all notified bodies designated under the AI Act within the European Union.
Recital (38) Markdown
European Union
Article 39 — Conformity assessment bodies of third countries Markdown
This article applies to conformity assessment bodies established in countries outside the European Union.
Recital (39) Markdown
This recital applies to all EU Member States as part of the harmonized framework for AI regulation.
Article 40 — Harmonised standards and standardisation activities Markdown
This article applies to all AI systems within the scope of the EU AI Act, facilitating compliance through standardised technical specifications.
Recital (40) Markdown
Ireland is excluded from the application of Article 5(1)(d), Article 26, and Article 5(2)-(6) of the AI Act under specific conditions related to its non-participation in certain EU police and judicial cooperation frameworks.
Article 41 — Common specifications Markdown
This article applies to all high-risk AI systems within the scope of the EU AI Act, regardless of the Member State, provided the conditions for the lack of harmonized standards are met.
Recital (41) Markdown
Denmark is excluded from the application of specific provisions of the AI Act related to law enforcement and certain prohibited AI practices as outlined in the Protocol.
Article 42 — Presumption of conformity with certain requirements Markdown
This article applies to high-risk AI systems within the scope of the EU AI Act.
Recital (42) Markdown
European Union
Article 43 — Conformity Assessment Markdown
This article applies to all high-risk AI systems placed on the EU market or put into service within the European Union.
Recital (43) Markdown
This regulation applies to all AI systems placed on the market or put into service within the European Union.
Article 44 — Certificates Markdown
This article applies to all EU Member States under the AI Act framework.
Recital (44) Markdown
This recital applies to the European Union under the AI Act framework.
Article 45 — Information obligations of notified bodies Markdown
This article applies to all notified bodies operating within the scope of the regulation, ensuring consistent communication with notifying authorities and peer organizations.
Recital (45) Markdown
European Union
Article 46 — Derogation from the conformity assessment procedure Markdown
This article applies to all EU Member States under the AI Act framework.
Recital (46) Markdown
This recital applies to all high-risk AI systems placed on the EU market, put into service, or used within the Union.
Article 47 — EU declaration of conformity Markdown
This article applies to providers of high-risk AI systems within the scope of the EU AI Act.
Recital (47) Markdown
This regulation applies to AI systems integrated into products within the European Union market.
Article 48 — CE Marking Markdown
The CE marking must be affixed to high-risk AI systems in accordance with the general principles set out in Regulation (EC) No 765/2008.
Recital (48) Markdown
This recital applies to all AI systems deployed within the European Union, as it defines the criteria for high-risk classification based on fundamental rights impact.
Article 49 — Registration Markdown
This article applies to high-risk AI systems placed on the EU market or put into service within the European Union.
Recital (49) Markdown
European Union
Article 50 - Article 50 — Transparency obligations for providers and deployers of certain AI systems Markdown
This article applies to providers and deployers of AI systems operating within the scope of the EU AI Act.
Recital (50) Markdown
This applies to AI systems integrated into products regulated under EU harmonization legislation, such as machinery, toys, medical devices, and vehicles, where third-party conformity assessment is mandatory.
Article 51 - Article 51 — Classification of general-purpose AI models as general-purpose AI models with systemic risk Markdown
Article 51 defines the criteria for classifying general-purpose AI models as having systemic risk, including high-impact capabilities and computational thresholds. The Commission may amend these thresholds and criteria.
Recital (51) Markdown
European Union
Article 52 - Article 52 — Procedure Markdown
This article applies to providers of general-purpose AI models that meet the criteria for systemic risk as defined under the AI Act.
Recital (52) Markdown
This recital applies to all EU Member States as part of the EU AI Act framework.
Article 53 - Article 53 — Obligations for providers of general-purpose AI models Markdown
Article 53 details the obligations for providers of general-purpose AI models, including technical documentation, copyright compliance, and cooperation with authorities. Certain obligations do not apply to models released under free and open-source licenses.
Recital (53) Markdown
This recital applies to the European Union under the AI Act framework.
Article 54 - Article 54 — Authorized representatives of providers of general-purpose AI models Markdown
This article applies to providers of general-purpose AI models established outside the European Union who place their models on the EU market.
Recital (54) Markdown
This recital applies to all EU Member States as part of the EU AI Act framework.
Article 55 - Article 55 — Obligations for providers of general-purpose AI models with systemic risk Markdown
This article applies to providers of general-purpose AI models with systemic risk operating within the European Union.
Recital (55) Markdown
This recital applies to AI systems used as safety components in critical infrastructure across the European Union.
Article 56 - Article 56 — Codes of practice Markdown
This regulation applies to the European Union.
Recital (56) Markdown
This recital applies to the European Union under the AI Act framework.
Article 57 — AI Regulatory Sandboxes Markdown
Article 57 establishes a framework for AI regulatory sandboxes, aimed at fostering innovation while ensuring regulatory compliance. Member States must establish controlled environments to test and validate AI systems before they are placed on the market.
Recital (57) Markdown
This recital applies to all EU Member States as part of the EU AI Act regulatory framework.
Article 58 — Detailed arrangements for AI regulatory sandboxes and their operation Markdown
This article applies to the European Union, governing the establishment and operation of AI regulatory sandboxes by national competent authorities.
Recital (58) Markdown
This recital applies to the European Union under the AI Act framework.
Article 59 — Further processing of personal data for the development of certain AI systems in the public interest Markdown
This article applies to the processing of personal data within AI regulatory sandboxes established under the EU AI Act.
Recital (59) Markdown
This recital clarifies the classification of AI systems used by law enforcement as high-risk due to their potential impact on fundamental rights, procedural fairness, and the presumption of innocence.
Article 60 — Testing of high-risk AI systems in real-world conditions outside AI regulatory sandboxes Markdown
This article applies to providers of high-risk AI systems intending to conduct real-world testing within the European Union.
Recital (60) Markdown
This recital applies to all EU Member States implementing AI systems within the scope of migration, asylum, and border management.
Article 61 — Informed consent to participate in AI real-world testing outside AI regulatory sandboxes Markdown
This article applies to providers or deployers conducting real-world testing of AI systems outside of regulatory sandboxes within the scope of the EU AI Act.
Recital (61) Markdown
This recital applies to all EU Member States as part of the EU AI Act framework.
Article 62 — Measures in support of providers and deployers, in particular SMEs, including start-ups Markdown
This article applies to all providers and deployers within the European Union, with a specific focus on supporting SMEs and start-ups to ensure compliance with the AI Act.
Recital (62) Markdown
This recital applies to all EU Member States.
Article 63 — Derogations for specific operators Markdown
This article applies to micro-enterprises operating within the scope of the regulation, allowing for simplified quality management processes.
Recital (63) Markdown
European Union
Article 64 — AI Office Markdown
The AI Office is established within the European Commission to oversee AI expertise and implementation across the EU.
Recital (64) Markdown
This recital applies to all high-risk AI systems placed on the market or put into service within the European Union.
Article 65 — Establishment and structure of the European Artificial Intelligence Board Markdown
Article 65 establishes the European AI Board, its composition, members, subgroups, and operational framework.
Recital (65) Markdown
EU
Article 66 — Tasks of the AI Board Markdown
The AI Board acts as a central advisory body to ensure uniform implementation of the AI Act across the European Union.
Recital (66) Markdown
European Union
Article 67 — Advisory Forum Markdown
The Advisory Forum is established to provide technical expertise and stakeholder input to the AI Office and the Commission regarding the implementation of the AI Act.
Recital (67) Markdown
EU
Article 68 — Scientific Panel of Independent Experts Markdown
Article 68 establishes a scientific panel of independent experts to support the enforcement of the AI Act, composed of experts selected for their expertise and independence.
Recital (68) Markdown
This applies to all EU Member States as part of the AI Act framework.
Article 69 — Member States' access to the scientific panel Markdown
Member States are entitled to request expert support to assist in the enforcement of the regulation, with the European Commission responsible for establishing fee structures and coordinating access to ensure cost-effectiveness and timely support.
Recital (69) Markdown
European Union
Article 70 — Designation of national competent authorities and single points of contact Markdown
Member States must designate one or more national competent authorities, including a national supervisory authority, to ensure the application and implementation of this Regulation.
Recital (70) Markdown
This recital applies to all providers of AI systems within the scope of the EU AI Act operating within the European Union.
Article 71 — EU database for high-risk AI systems listed in Annex III Markdown
This regulation applies to all EU Member States.
Recital (71) Markdown
This recital applies to all high-risk AI systems within the scope of the EU AI Act.
Article 72 — Post-market monitoring by providers and post-market monitoring plan for high-risk AI systems Markdown
This article applies to providers of high-risk AI systems operating within the scope of the EU AI Act.
Recital (72) Markdown
This recital applies to all high-risk AI systems deployed within the European Union, as defined by the AI Act.
Article 73 — Reporting of serious incidents Markdown
This article applies to all providers of high-risk AI systems operating within the European Union.
Recital (73) Markdown
This recital applies to all high-risk AI systems within the scope of the EU AI Act, requiring providers to integrate human oversight mechanisms during the design and development phases.
Article 74 — Market surveillance and control of AI systems on the Union market Markdown
This article applies to all Member States of the European Union, governing the market surveillance of AI systems placed on the Union market.
Recital (74) Markdown
EU
Article 75 — Mutual assistance, market surveillance and control of general-purpose AI systems Markdown
This article applies to all EU Member States, mandating cooperation between national market surveillance authorities and the European AI Office.
Recital (75) Markdown
High-risk AI systems must be designed to be technically robust, ensuring resilience against harmful or undesirable behaviors through appropriate technical and organizational measures.
Article 76 — Supervision of real-world testing by market surveillance authorities Markdown
This article applies to market surveillance authorities within the European Union responsible for overseeing AI systems undergoing real-world testing.
Recital (76) Markdown
Cybersecurity is essential for ensuring AI system resilience against attempts to alter their use, behavior, or performance, or to compromise their security properties by exploiting vulnerabilities.
Article 77 — Powers of fundamental rights protection authorities Markdown
These provisions apply to designated fundamental rights protection authorities within the European Union member states as defined under the AI Act.
Recital (77) Markdown
This recital applies to high-risk AI systems subject to EU-wide cybersecurity regulations.
Article 78 — Confidentiality Markdown
This article applies to all entities subject to the EU AI Act, ensuring that information exchanged during regulatory processes remains protected.
Recital (78) Markdown
EU
Article 79 — Procedure applicable at national level to AI systems presenting a risk Markdown
This article applies to all EU Member States as part of the AI Act regulatory framework.
Recital (79) Markdown
This recital applies to all providers of high-risk AI systems placing products on the EU market, including those established outside the Union who must appoint an authorized representative.
Article 80 — Procedure applicable to AI systems classified by the provider as not high-risk pursuant to Annex III Markdown
This article applies to all AI systems placed on the Union market that are subject to the AI Act's classification requirements.
Recital (80) Markdown
The Union and Member States are bound by the UN Convention on the Rights of Persons with Disabilities to ensure equal access to AI systems.
Article 81 — Union safeguard procedure Markdown
This procedure applies to all EU Member States when a national measure taken against an AI system is challenged or when the Commission identifies potential non-compliance with the AI Act.
Recital (81) Markdown
This recital applies to all providers of high-risk AI systems within the scope of the EU AI Act.
Article 82 — Compliant AI systems which present a risk Markdown
Article 82 establishes the framework for actions taken when compliant AI systems present a risk, involving supervisory authorities, operators, and Member States.
Recital (82) Markdown
This applies to providers of AI systems established outside the European Union who place their systems on the Union market.
Article 83 — Formal non-compliance Markdown
This article applies to all high-risk AI systems placed on the EU market.
Recital (83) Markdown
EU
Article 84 — Union testing support structures for AI Markdown
This article applies to all EU Member States as part of the EU AI Act framework.
Recital (84) Markdown
This recital applies to all EU member states as part of the EU AI Act.
Article 85 — Right to lodge a complaint with a market surveillance authority Markdown
This article applies to all EU Member States under the AI Act framework.
Recital (85) Markdown
This recital applies to all EU Member States as part of the AI Act framework.
Article 86 — Right to an explanation of individual decisions Markdown
This article applies to deployments of high-risk AI systems within the scope of the EU AI Act.
Recital (86) Markdown
This applies to all economic operators within the scope of the EU AI Act who may assume the role of provider following a change in the status of the original provider.
Article 87 — Reporting of breaches and protection of reporting persons Markdown
This article applies to all entities and individuals subject to the Regulation within the European Union.
Recital (87) Markdown
This recital applies to high-risk AI systems integrated into products regulated under the EU New Legislative Framework.
Article 88 — Monitoring of the implementation of obligations for providers of general-purpose AI models Markdown
This article applies to the enforcement of obligations for providers of general-purpose AI models within the scope of the EU AI Act.
Recital (88) Markdown
This recital applies to all entities involved in the AI value chain within the scope of the EU AI Act.
Article 89 — Control measures Markdown
Article 89 establishes the oversight powers of the AI Office and provides a formal mechanism for downstream providers to report non-compliance by AI model providers.
Recital (89) Markdown
This recital applies to the European Union regulatory framework for Artificial Intelligence.
Article 90 — Alerts on systemic risks issued by the scientific panel Markdown
This article applies to the European Union regulatory framework for AI models.
Recital (90) Markdown
The Commission may develop and recommend voluntary model contractual clauses to be established between providers of high-risk AI systems and third parties providing tools, services, components, or processes used in or integrated into high-risk AI systems to facilitate cooperation throughout the AI value chain.
Article 91 — Power to request documentation and information Markdown
Article 91 outlines the Commission's power to request information from AI model providers, including the procedures and responsibilities involved.
Recital (91) Markdown
This recital applies to all EU Member States as part of the AI Act framework.
Article 92 — Power to conduct evaluations Markdown
This article applies to all providers of AI models subject to the AI Act within the European Union.
Recital (92) Markdown
The regulation ensures that existing labor rights regarding the information and consultation of workers remain unaffected by the deployment of AI systems.
Article 93 — Power to request measures Markdown
This article applies to providers of AI models within the scope of the EU AI Act.
Recital (93) Markdown
This recital applies to the deployment of high-risk AI systems within the European Union, emphasizing the responsibilities of deployers in mitigating risks and ensuring transparency for natural persons.
Article 94 — Procedural rights of providers of general-purpose AI models Markdown
Article 94 extends the procedural rights of providers of general-purpose AI models, in accordance with Regulation (EU) 2019/1020, while taking into account the specific rights provided for by this Regulation.
Recital (94) Markdown
This recital applies to the processing of biometric data by AI systems used for law enforcement purposes within the European Union.
Article 95 - Article 95 — Codes of conduct for voluntary application of certain requirements Markdown
Article 95 encourages voluntary codes of conduct for AI, excluding high-risk systems, while taking into account best practices and the specific needs of SMEs.
Recital (95) Markdown
This provision applies to all EU Member States as part of the harmonized regulatory framework for Artificial Intelligence.
Article 96 — Commission guidelines on the implementation of this Regulation Markdown
This article applies to the European Union, providing a framework for the Commission to issue guidance on the interpretation and application of the AI Act.
Recital (96) Markdown
This requirement applies to deployers of high-risk AI systems within the European Union, specifically public bodies and private entities providing public services, as well as those specified in Annex III, point 5(b).
Article 97 — Exercise of the delegation Markdown
Article 97 defines the conditions for the Commission's exercise of delegated powers, subject to strict oversight by the European Parliament and the Council.
Recital (97) Markdown
The concept of general-purpose AI models should be clearly defined and distinct from AI systems to ensure legal certainty. The definition should be based on the...
Article 98 — Committee Markdown
Article 98 establishes a committee to assist the Commission, in accordance with Regulation (EU) No 182/2011, with the application of Article 5 for reference.
Recital (98) Markdown
The criteria for model generality under the AI Act include, but are not limited to, a threshold of at least one billion parameters combined with training on large-scale data sets using extensive supervision.
Article 99 - Article 99 — Penalties Markdown
Article 99 of the AI Act establishes the penalty regime for violations of AI rules, with proportionate and dissuasive fines adapted for SMEs. Member States must inform the Commission of their penalty regime and apply graduated fines according to the severity of the infringements.
Recital (99) Markdown
This recital clarifies the classification of large generative AI models as general-purpose AI models under the EU AI Act.
Article 100 — Administrative fines imposed on Union institutions, bodies, offices and agencies Markdown
This article applies to all Union institutions, bodies, offices, and agencies subject to the AI Act.
Recital (100) Markdown
This recital clarifies the classification of AI systems that incorporate general-purpose AI models, establishing that the integration of such models confers general-purpose status if it enables the system to perform a broad range of distinct tasks.
Article 101 — Fines applicable to providers of general-purpose AI models Markdown
This regulation applies to providers of general-purpose AI models operating within the European Union or serving the EU market.
Recital (101) Markdown
This recital applies to providers of general-purpose AI models operating within the scope of the EU AI Act.
Article 102 — Amendment to Regulation (EC) No 300/2008 Markdown
Article 102 amends Regulation (EC) No 300/2008 to include technical specifications and approval procedures for AI-based security equipment, in accordance with Regulation (EU) 2024/1689.
Recital (102) Markdown
This recital applies to all AI systems and models developed or deployed within the scope of the EU AI Act, regardless of the specific geographic location of the provider, provided they are placed on the EU market or put into service in the EU.
Article 103 — Amendment to Regulation (EU) No 167/2013 Markdown
Article 103 amends Regulation (EU) No 167/2013 by adding a subparagraph to Article 17(5) concerning AI systems used as safety components.
Recital (103) Markdown
The regulation defines free and open-source AI components to include software, data, models, and tools, provided they are not monetized or provided through open repositories.
Article 104 — Amendment to Regulation (EU) No 168/2013 Markdown
Article 104 amends Regulation (EU) No 168/2013 by adding a subparagraph to Article 22(5) concerning AI systems as safety components.
Recital (104) Markdown
This recital applies to all providers of general-purpose AI models operating within the European Union, specifically addressing those utilizing free and open-source licensing frameworks.
Article 105 — Amendment to Directive 2014/90/EU Markdown
Article 105 amends Directive 2014/90/EU to include specific requirements for AI systems used as safety components, in accordance with Regulation (EU) 2024/1689.
Recital (105) Markdown
This recital applies to the development and training of general-purpose AI models within the European Union, specifically addressing the intersection of AI development and copyright law regarding text and data mining.
Article 106 — Amendment to Directive (EU) 2016/797 Markdown
Article 106 amends Directive (EU) 2016/797 by adding a paragraph concerning AI systems considered as safety components.
Recital (106) Markdown
This applies to all providers of general-purpose AI models placing their products on the Union market, regardless of their place of establishment.
Article 107 — Amendment to Regulation (EU) 2018/858 Markdown
This article applies to the European Union regulatory framework concerning the type-approval of motor vehicles.
Recital (107) Markdown
This recital applies to all providers of general-purpose AI models operating within the scope of the EU AI Act.
Article 108 — Amendments to Regulation (EU) 2018/1139 Markdown
This regulation applies to the European Union.
Recital (108) Markdown
This recital applies to all providers of general-purpose AI models operating within the European Union.
Article 109 — Amendment to Regulation (EU) 2019/2144 Markdown
This regulation applies to the European Union.
Recital (109) Markdown
The regulation applies to providers of general-purpose AI models, with obligations scaled based on the provider's nature, the model's systemic risk, and the extent of modifications made.
Article 110 — Amendment to Directive (EU) 2020/1828 Markdown
Article 110 amends Directive (EU) 2020/1828 by adding a point to its Annex I regarding the Artificial Intelligence Act.
Recital (110) Markdown
This recital applies to the European Union under the AI Act framework.
Article 111 — AI systems already placed on the market or put into service and general-purpose AI models already placed on the market Markdown
This article applies to all AI systems and general-purpose AI models already placed on the market or put into service within the European Union.
Recital (111) Markdown
This recital applies to the classification of general-purpose AI models within the European Union under the AI Act.
Article 112 — Evaluation and review Markdown
This article applies to the European Commission's oversight and review process of the AI Act across all EU Member States.
Recital (112) Markdown
This recital applies to all providers of general-purpose AI models within the scope of the EU AI Act.
Article 113 - Article 113 — Entry into force and application Markdown
The regulation applies across all EU Member States.
Recital (113) Markdown
This recital applies to all general-purpose AI models operating within the scope of the EU AI Act, regardless of the provider's location, provided the model is placed on the EU market or its output is used in the EU.
Recital (114) Markdown
This recital applies to providers of general-purpose AI models with systemic risks operating within the European Union.
Recital (115) Markdown
This recital applies to providers of general-purpose AI models that are classified as having systemic risks under the EU AI Act.
Recital (116) Markdown
This recital applies to all EU Member States as part of the harmonized regulatory framework for AI.
Recital (117) Markdown
This recital applies to all providers of general-purpose AI models operating within the European Union.
Recital (118) Markdown
The regulation applies to AI systems and models within the European Union, ensuring that AI integrated into very large online platforms adheres to both the AI Act and the risk management framework of Regulation (EU) 2022/2065.
Recital (119) Markdown
This recital emphasizes the need for a technology-neutral interpretation of Union law to ensure regulatory consistency amidst rapid digital innovation.
Recital (120) Markdown
This recital applies to all EU Member States as part of the implementation of the AI Act in conjunction with the Digital Services Act (Regulation (EU) 2022/2065).
Recital (121) Markdown
This recital applies to all providers subject to the AI Act within the European Union, emphasizing the role of harmonized standards and common specifications in achieving regulatory compliance.
Recital (122) Markdown
This recital applies to all EU Member States under the scope of the AI Act.
Recital (123) Markdown
This applies to all high-risk AI systems within the scope of the EU AI Act.
Recital (124) Markdown
This recital applies to high-risk AI systems integrated into products subject to existing Union harmonization legislation, ensuring that conformity assessments are streamlined to avoid redundant procedures.
Recital (125) Markdown
This recital applies to all EU Member States as part of the AI Act framework.
Recital (126) Markdown
The regulation requires national competent authorities to notify third-party conformity assessment bodies to the Commission and other Member States, ensuring they meet strict regulatory requirements.
Recital (127) Markdown
The European Union
Recital (128) Markdown
This recital applies to all high-risk AI systems deployed within the European Union market.
Recital (129) Markdown
The 'CE' marking serves as a declaration of conformity for high-risk AI systems, allowing for free movement within the EU internal market. It must be affixed physically for integrated systems or digitally for software-only systems, in accordance with Regulation (EC) No 765/2008.
Recital (130) Markdown
The regulation allows for the exceptional deployment of AI systems without prior conformity assessment in cases of urgent public interest, such as public safety, health, or environmental protection, subject to oversight by market surveillance authorities.
Recital (131) Markdown
The EU database for high-risk AI systems is designed to increase transparency and facilitate oversight by the Commission and Member States, with specific registration requirements for providers and certain public deployers.
Recital (132) Markdown
This recital applies to all AI systems interacting with natural persons or generating content within the scope of the EU AI Act.
Recital (133) Markdown
This recital applies to all providers of AI systems generating synthetic content within the scope of the EU AI Act.
Recital (134) Markdown
In addition to the technical solutions used by system providers, deployers who use an AI system to generate or manipulate images or audio or video content presenting...
Recital (135) Markdown
This recital applies to all EU Member States as part of the EU AI Act framework.
Recital (136) Markdown
This recital applies to all EU Member States as part of the EU AI Act.
Recital (137) Markdown
European Union
Recital (138) Markdown
The regulation mandates that Member States establish at least one AI regulatory sandbox at the national level to provide a controlled environment for testing innovative AI systems under the supervision of competent authorities.
Recital (139) Markdown
This recital applies to all Member States of the European Union as part of the AI Act framework.
Recital (140) Markdown
This Regulation should provide the legal basis for the use of personal data by providers and prospective providers in the AI regulatory sandbox for developing innovative AI systems.
Recital (141) Markdown
This recital applies to all EU Member States as part of the EU AI Act framework governing the testing of high-risk AI systems in real-world conditions.
Recital (142) Markdown
European Union
Recital (143) Markdown
This recital applies to all EU Member States and entities operating within the scope of the AI Act.
Recital (144) Markdown
European Union
Recital (145) Markdown
This recital applies to all EU Member States as part of the AI Act framework.
Recital (146) Markdown
This recital applies to all microenterprises operating within the scope of the AI Act, regardless of their specific geographic location within the EU.
Recital (147) Markdown
This provision applies to the European Union, specifically targeting entities established or accredited under Union law.
Recital (148) Markdown
This recital applies to all EU Member States as part of the AI Act governance structure.
Recital (149) Markdown
This recital applies to all Member States of the European Union as it pertains to the governance structure of the AI Act.
Recital (150) Markdown
This recital applies to the European Union as part of the AI Act framework.
Recital (151) Markdown
The scientific panel is established to provide independent expertise to the AI Office and Member States for the implementation and enforcement of the AI Act, specifically concerning general-purpose AI models.
Recital (152) Markdown
This recital applies to all EU Member States as part of the AI Act regulatory framework.
Recital (153) Markdown
Member States are responsible for designating notifying authorities and market surveillance authorities to ensure compliance with the Regulation.
Recital (154) Markdown
This recital applies to all EU Member States designated as national competent authorities under the Regulation.
Recital (155) Markdown
This recital applies to all providers of high-risk AI systems within the scope of the EU AI Act.
Recital (156) Markdown
This recital applies to all EU Member States and the institutions, agencies, and bodies of the Union.
Recital (157) Markdown
This recital applies to all EU Member States and their respective national public authorities responsible for supervising fundamental rights and AI compliance.
Recital (158) Markdown
The regulation applies to financial institutions across the European Union, requiring coordination between AI supervisory authorities and financial regulators like the ECB.
Recital (159) Markdown
This applies to all EU Member States where market surveillance authorities oversee high-risk AI systems used by law enforcement.
Recital (160) Markdown
This recital applies to all EU Member States under the AI Act framework.
Recital (161) Markdown
The AI Office oversees AI systems and models provided by the same provider, while national authorities retain oversight for other AI systems, with cooperation mechanisms in place for high-risk systems.
Recital (162) Markdown
The European Commission, through the AI Office, is responsible for the centralized supervision and enforcement of obligations for providers of general-purpose AI models with systemic risk.
Recital (163) Markdown
This recital applies to the governance framework for general-purpose AI models within the European Union under the AI Act.
Recital (164) Markdown
This recital applies to the oversight and enforcement powers of the AI Office regarding providers of general-purpose AI models within the scope of the EU AI Act.
Recital (165) Markdown
This recital applies to all EU Member States as part of the AI Act framework.
Recital (166) Markdown
This recital applies to the European Union jurisdiction as part of the EU AI Act.
Recital (167) Markdown
This recital applies to all EU Member States and relevant Union-level competent authorities involved in the enforcement of the Regulation.
Recital (168) Markdown
This recital applies to all EU Member States subject to the AI Act.
Recital (169) Markdown
This recital clarifies that providers of general-purpose AI models are subject to enforcement measures, including fines, for non-compliance with the AI Act.
Recital (170) Markdown
The regulation ensures that existing legal avenues for redress remain applicable to AI-related infringements.
Recital (171) Markdown
This provision applies to all EU Member States under the AI Act framework.
Recital (172) Markdown
This recital applies to all EU Member States as part of the implementation of the AI Act.
Recital (173) Markdown
This recital applies to the European Union and all Member States subject to the AI Act.
Recital (174) Markdown
This recital applies to the European Union as part of the AI Act regulatory framework.
Recital (175) Markdown
This recital establishes the framework for the Commission's implementing powers to ensure consistent application of the Regulation across Member States, in line with the examination procedure set out in Regulation (EU) No 182/2011.
Recital (176) Markdown
This recital justifies the EU's competence to legislate on AI by invoking the principles of subsidiarity and proportionality, arguing that the objectives of the AI Act cannot be effectively achieved by Member States individually.
Recital (177) Markdown
This recital applies to all EU Member States subject to the AI Act.
Recital (178) Markdown
This recital applies to all providers of high-risk AI systems within the scope of the EU AI Act.
Recital (179) Markdown
The AI Act generally applies from 2 August 2026, with specific staggered implementation dates for prohibitions (2 February 2025), governance and penalties (2 August 2025), and general-purpose AI model obligations (2 August 2025).
Recital (180) Markdown
European Union