Entry Into Force

Legal Framework

Article 113 of the AI Act establishes a staggered entry-into-force regime that creates distinct compliance deadlines for different categories of obligations. The regulation enters into force on the twentieth day following publication in the Official Journal of the European Union, but its substantive provisions become applicable on a phased schedule. Chapters I (General Provisions) and II (Prohibited AI Practices) apply from February 2, 2025. Chapter III Section 4 (Notifying Authorities), Chapter V (Governance), Chapter VII (Post-Market Monitoring), Chapter XII (Codes of Conduct), and Article 78 apply from August 2, 2025, with the exception of Article 101. The core classification obligation under Article 6(1) and its corresponding requirements apply from August 2, 2027. The general applicability date for remaining provisions is August 2, 2026.

This phased approach mirrors the GDPR model under Article 99, where the regulation entered into force on May 25, 2016, but became applicable two years later on May 25, 2018. During that transitional period, controllers were required to bring ongoing processing operations into compliance, though consent previously obtained under the Data Protection Directive did not need to be re-obtained.

Key Developments

Dutch case law establishes clear principles on temporal application of new legislation. The WAMCA ruling clarifies that past events are not assessed under new law when a party could not have anticipated it, but conduct initiated before entry into force that continues afterward falls within the new regime — the party is deemed to have accepted exposure by continuing the behavior. The Woo jurisprudence reinforces this: where no transitional provisions are enacted, new legislation takes immediate effect and applies to all decisions rendered after the entry-into-force date, regardless of when the underlying request was filed. Courts have consistently applied this principle — a decision dated after entry into force is governed by the new legal framework even if the procedural history predates it.

The absence of transitional law means organizations cannot rely on pending procedures or pre-existing arrangements to defer compliance with newly applicable provisions.

Practical Guidance

• Map obligations to phased deadlines immediately: Identify which AI Act provisions govern your systems and categorize compliance work against the three key dates — February 2, 2025 (prohibited practices), August 2, 2025 (governance and post-market monitoring), and August 2, 2026 (general applicability), with high-risk classification obligations under Article 6(1) extending to August 2, 2027.

• Audit ongoing conduct against the continuation principle: Under the WAMCA standard, AI systems deployed before a relevant entry-into-force date but continuing in operation afterward will be subject to the new regime. Document when each system was placed on the market and whether it remains in use.

• Prepare for immediate application without transitional relief: The Woo cases confirm that absent transitional provisions, new law applies to all decisions and actions taken after the applicability date. Do not assume pending assessments or pre-existing compliance frameworks will grandfather your operations.

• Use the pre-applicability period substantively: Following the GDPR model, the period between entry into force and applicability is intended for operational alignment, not mere awareness. Begin gap assessments, documentation, and governance restructuring now.

• Reassess consent and lawful bases: Where AI systems process personal data, verify that existing consent obtained under prior legal frameworks remains valid under current GDPR standards, as re-obtaining consent is not automatically required but adequacy must be confirmed.