EU AI Act — Articles

AI systems posing unacceptable risk that are banned in the EU, including social scoring, manipulative techniques, and biometric categorization.

• Art. 5Article 5 — Prohibited AI practices

AI systems requiring strict compliance obligations, including risk management, data quality, transparency, and human oversight.

• Art. 10Article 10 — Data and data governance
• Art. 11Article 11 — Technical documentation
• Art. 12Article 12 — Record-keeping
• Art. 13Article 13 — Transparency and provision of information to deployers
• Art. 14Article 14 — Human oversight
• Art. 15Article 15 — Accuracy, robustness and cybersecurity
• Art. 16Article 16 — Obligations of providers of high-risk AI systems
• Art. 17Article 17 — Quality management system
• Art. 18Article 18 — Documentation keeping
• Art. 19Article 19 — Automatically generated logs
• Art. 20Article 20 — Corrective actions and duty of information
• Art. 21Article 21 — Cooperation with competent authorities
• Art. 22Article 22 — Authorised representatives of providers of high-risk AI systems
• Art. 23Article 23 — Obligations of importers
• Art. 24Article 24 — Obligations of distributors
• Art. 25Article 25 — Responsibilities along the AI value chain
• Art. 26Article 26 — Obligations of deployers of high-risk AI systems
• Art. 27Article 27 — Fundamental rights impact assessment for high-risk AI systems
• Art. 6Article 6 — Classification rules for high-risk AI systems
• Art. 7Article 7 — Amendments to Annex III
• Art. 8Article 8 — Compliance with the requirements
• Art. 9Article 9 — Risk management system

AI systems with transparency obligations — users must know they are interacting with AI, including chatbots, deepfakes, and generative AI.

• Art. 50Article 50 — Transparency obligations for providers and deployers of certain AI systems

General-purpose models like large language models (LLMs) with specific transparency and systemic risk assessment obligations.

Institutional framework for enforcing the EU AI Act, including the AI Office, national authorities, sandboxes, and penalties.