Guidelines on prohibited AI practices
These guidelines explain how the prohibitions in Article 5 of the AI Act should be interpreted and applied. They cover harmful manipulation and deception, exploitation of vulnerabilities, social scoring, individual criminal offence risk assessments relying solely on profiling or personality characteristics, untargeted scraping of facial images, emotion recognition in workplaces and educational institutions, biometric categorisation involving sensitive characteristics, and real-time remote biometric identification for law enforcement purposes. They also clarify the responsible actors, scope, exceptions, and interaction with other Union legislation through practical examples.
Particular attention is given to the limited exceptions and safeguards for real-time remote biometric identification, including necessity, proportionality, prior authorisation, registration and notification requirements.
Application date: The prohibitions in Article 5 covered by these guidelines have applied since 2 February 2025. The new prohibitions introduced through the Digital Omnibus will apply from 2 December 2026, but they are not covered by the guidelines published in February 2025.
Information: Guidelines on prohibited artificial intelligence practices established by Regulation (EU) 2024/1689 (AI Act)
Guidelines in Latvian: Komisijas Pamatnostādnes par aizliegtu mākslīgā intelekta praksi, kas noteikta Regulā (ES) 2024/1689 (MI aktā)
Guidelines in English: Commission Guidelines on prohibited artificial intelligence practices established by Regulation (EU) 2024/1689 (AI Act)
Guidelines on the definition of an AI system
These guidelines explain the definition in Article 3(1) and help determine whether a software system falls within the scope of the AI Act. They analyse the seven elements of the definition: a machine-based system, varying levels of autonomy, possible adaptiveness after deployment, explicit or implicit objectives, inference from inputs, generation of outputs, and the ability of those outputs to influence physical or virtual environments. They also provide examples of systems that may fall outside the definition, including basic data processing, conventional mathematical optimisation, and systems based solely on rules defined by natural persons.
Application date: The definition of an AI system in Article 3(1) has applied since 2 February 2025, together with the other provisions of Chapter I of the AI Act.
Guidelines in Latvian: Komisijas pamatnostādnes par mākslīgā intelekta sistēmas definīciju, kas noteikta Regulā (ES) 2024/1689 (MI aktā)
Guidelines in English: Guidelines on the definition of an artificial intelligence system established by Regulation (EU) 20241689 (AI Act)
Guidelines on obligations for providers of general-purpose AI models
These guidelines clarify the scope of Chapter V of the AI Act, particularly Articles 51 to 55. They explain what constitutes a general-purpose AI model, when an entity qualifies as its provider, and when modification or further training may transfer provider obligations to another entity. They also address open-source exemptions, models with systemic risk, computational thresholds, placing models on the EU market, cooperation throughout the AI value chain, and the application of transparency, copyright, documentation, risk management, and incident reporting obligations.
Application date: The obligations for providers of general-purpose AI models have applied since 2 August 2025 to models placed on the market from that date. Providers of models placed on the market before 2 August 2025 must comply by 2 August 2027.
Information: Guidelines on the scope of obligations for providers of general-purpose AI models under the AI Act
Guidelines in Latvian: Komisijas pamatnostādnes par vispārīga lietojuma MI modeļu nodrošinātāju pienākumu tvērumu saskaņā ar Regulu (ES) 2024/1689 (MI akts)
Guidelines in English: Commission Guidelines on the scope of the obligations for providers of general-purpose AI models established by Regulation (EU) 2024/1689 (AI Act)
Guidelines on transparency obligations for providers and deployers of AI systems
These guidelines explain Article 50 obligations applying to providers and deployers of systems that interact directly with natural persons, generate or manipulate content, perform emotion recognition or biometric categorisation, or produce deep fakes and certain public interest text. They clarify when people must be informed that they are interacting with an AI system, how synthetic content must be marked in a machine readable and detectable format, and when generated or manipulated content must be visibly disclosed. They also address exceptions, timing, accessibility, technical feasibility, and the respective responsibilities of providers and deployers.
Application date: The transparency obligations in Article 50 apply from 2 August 2026. Providers of systems generating synthetic audio, images, video or text that were placed on the market before that date must comply with the marking obligation in Article 50(2) by 2 December 2026.
Information: Guidelines on transparency obligations for providers and deployers of AI systems
Guidelines in English: Guidelines on the implementation of the transparency obligations for certain AI systems under Article 50 of Regulation (EU) 2024/1689 (AI Act)