European Union AI Act Published In The Official Journal—Critical Milestones On The Road To Full Applicability

August 1, 2024

Entry into force of the AI Act (Article 113). This means that the AI Act becomes part of the EU legal order. It does not mean that the provisions of the AI Act are applicable.

November 2, 2024

By this date, Member States must have identified the public authorities or bodies that supervise or enforce obligations under EU law protecting fundamental rights, including the right to nondiscrimination, in relation to the use of high-risk AI systems referred to in Annex III of the AI Act (Article 77(2)).

February 2, 2025

Chapters I and II will apply from this date (Article 113(a)). These include the general provisions (e.g., geographic scope, definitions) and the provisions on prohibited AI practices (see our client alert). Therefore, evaluating activities that might be considered "prohibited practices" under Article 5 of the AI Act should start as soon as possible.

The general obligation to ensure a sufficient level of AI literacy of staff under Article 4 of the AI Act will also apply from this date.

May 2, 2025

By this date, codes of practice for the implementation of general-purpose AI models and related obligations must be ready (Article 56(9)). These codes should support providers in achieving compliance with their duties relating to general-purpose AI models.

August 2, 2025

From this date, Chapter III, Section 4 (Notifying authorities and notified bodies), Chapter V (General-purpose AI models), Chapter VII (Governance) and Chapter XII (Penalties) will apply (except for Article 101, which deals with fines for providers of general-purpose AI models).

• Chapter III, Section 4 deals with notifying authorities and notified bodies, which are essential for the establishment of conformity assessment bodies.
• Chapter V contains the provisions related to general-purpose AI models introduced late in the legislative process; for example, the mandatory notification procedure for the provider (Article 52 (1)), documentation requirements (Article 53) and the appointment of an authorized representative (Article 54). Article 55 contains additional responsibilities focusing on the evaluation and mitigation of systemic risk and cyber and infrastructure security.
• Chapter VII sets out the governance structure on the EU level, including the AI Office, the European Artificial Intelligence Board, the advisory forum and the scientific panel. On the Member State level, the competent authorities must be appointed by this date (Article 70(2)).

• Chapter XII deals with penalties. This includes Article 99(3), which specifies the fines for noncompliance with prohibited AI practices referred to in Article 5. These fines can reach €35 million, or up to 7% of worldwide annual revenue, if the offender is an undertaking.

February 2, 2026

By this date, the European Commission will issue implementing acts creating a template for high-risk AI providers' post-market monitoring plans, which should serve as the basis for said monitoring system established by Article 72.

Similarly, the European Commission must, by this date, provide guidelines for the practical implementation of Article 6 concerning the classification of an AI system as high risk (Article 6(5)).

August 2, 2026

This is the default date for the provisions of the AI Act being applicable.

The obligations regarding high-risk AI systems will apply from this date, including risk and quality management systems, diligent data governance, technical documentation, recordkeeping, and transparency and information obligations.

Chapter IV addresses operators of AI systems directly interacting with humans, generative AI systems, and emotion recognition or biometric categorization systems, introducing disclosure and information responsibilities (see our client alert).

By this date, Member States must have implemented rules on penalties and other enforcement measures and notified the European Commission about them (Article 99).

At least one AI regulatory sandbox must be operational at a national level (Article 57(1)).

August 2, 2027

This is the ultimate deadline for AI systems covered by existing harmonization legislation (Article 113(c)) and for providers of general-purpose AI models that have been placed on the market up to 12 months after August 1, 2024, to comply with the AI Act.