EU AI Act compliance deadlines: the complete timeline

Regulation (EU) 2024/1689 enters into force

In force

The EU AI Act was published in the Official Journal of the European Union and entered into force on 1 August 2024. The Act does not apply in full from this date — most obligations take effect on later application dates. However, from this date the regulatory framework is established and businesses should begin planning for compliance.

Article 5 — Prohibited AI practices apply

In force since Feb 2025

The first substantive obligations: a list of AI practices is banned outright. These include subliminal manipulation that affects a person's behaviour against their interests, exploitation of vulnerabilities of protected groups, social scoring systems operated by public authorities, most forms of real-time remote biometric identification in publicly accessible spaces, biometric categorisation to infer sensitive attributes, emotion recognition in workplace and educational settings, and AI-assisted predictive policing based solely on profiling.

Full guide to Article 5 prohibited practices

General-purpose AI model (GPAI) obligations — Title IV

In force

Obligations for providers of general-purpose AI models (such as large language models distributed to third parties) apply from this date. These obligations — including transparency documentation, technical documentation, and copyright policy — primarily affect AI model developers and distributors, not SMEs that use commercial AI tools as deployers.

Article 4 (AI literacy) and Article 50 (transparency) apply

In force now

This is the most significant date for SMEs. Two new obligations now apply to all deployers of AI systems:

• Article 4 (AI literacy): Take proportionate measures to ensure staff who interact with AI systems have sufficient AI literacy. No specific certification format is mandated.
• Article 50 (transparency): Inform users when they are interacting with an AI system (unless obvious from context). Apply technical disclosure measures to AI-generated synthetic content.

Read the full Article 4 guide

Read the full Article 50 guide

Two new Article 5 prohibitions — Regulation (EU) 2026/1744

Upcoming

Regulation (EU) 2026/1744 (the Omnibus amendment) extended the Article 5 prohibited practices list with two additional prohibitions that apply from this date. Review the official EUR-Lex text for their exact scope.

Businesses should review these new prohibitions against their current and planned AI deployments and ensure no activity falls within the expanded prohibited list before 2 December 2026.

Chapter III — Annex III high-risk AI full obligations

Amended from Aug 2026

The original application date for Chapter III (Annex III high-risk AI) obligations was 2 August 2026. Regulation (EU) 2026/1744 amended this to 2 December 2027, giving businesses with high-risk AI use-cases more preparation time.

The heaviest obligations under the EU AI Act — covering AI systems in the Annex III high-risk categories — apply from this date. For deployers, the key obligation is Article 26: fundamental rights impact assessments, human oversight measures, EU database registration, and cooperation with supervisory authorities.

High-risk categories include: employment and HR AI (CV screening, performance assessment), credit and insurance scoring, educational assessment AI, critical infrastructure management, and several others.

Full guide to Annex III high-risk AI

Annex I regulated-product high-risk AI systems

Future

AI systems embedded in regulated products already covered by EU product safety law — such as medical devices, machinery, and aviation equipment (Annex I) — are subject to the full Chapter III obligations from this date. This is relevant primarily to manufacturers of regulated products and their supply chains.