EU AI Act: Timelines
Thursday August 1, 2024 is the date of entry into force of the EU AI Act
The AI revolution is bifurcating into seasonal pessimism and emerging practical applications. I’ll look at the pessimism in more detail later. Today, some practicality: we now know the date of entry into force of the EU AI Act.
AI Act Timelines
Following its adoption by Parliament and its publication in the Official Journal of the European Union the “Regulation laying down harmonised rules on Artificial Intelligence” (AI Act) will soon be live. Timelines and applicability are blooming in many corners of the AI legal space. In very simple terms, the higher the risk assessed by the EU, the sooner obligations on AI developers, providers, users, representatives, importers, and exporters will be effective. Let’s summarize everything to get our heads straight.
2024 – 01 August – Entry into Force of the AI Act
The AI Act was published in the Official Journal of the European Union on 12 July 2024 and so will come into force twenty days later on 01 August 2024.
2024 – 02 November – Certain Public Authorities Identified
National authorities will identify their public authorities or bodies that can request and access any relevant documentation in relation to high-risk AI systems’ respect of fundamental rights, including non-discrimination, under Art. 77(2).
2025 – 02 February – General AI Literacy and Initial Prohibitions
Provisions on AI literacy requirements and prohibited practices will apply. The prohibited practices include bans on AI systems that pose unacceptable risks, such as emotion recognition in workplaces, social scoring, predictive policing, and biometric categorisation based on sensitive characteristics. From Art. 113(a), this will be:
Chapter I (“General Provisions” including AI literacy) and Chapter II (prohibitions on unacceptable risk)
2025 – 02 February – “Codes of Practice”
Codes of practice will be available from the AI Office by this date, at the latest (with additional conditions including EU common rules if those codes aren’t available by 02 Aug 2025) under Art. 56(9).
2025 – 02 August – “Codes of Conduct” for General Purpose AI (GPAI)
The publication and application of codes of conduct for GPAI will begin, along with governance rules and penalties for new GPAI systems. Chapters and articles related to notifying authorities, governance, and sanctions will become applicable, though sanctions for breaches will not be enforced until the general applicability of these rules. From Art. 70(2), 73(7), and Art. 113(b), this will be:
Member States will make publicly available information on how competent authorities and single points of contact can be contacted
Commission guidance available for “serious incident” reports to the surveillance authorities
Chapter III, Section 4 (“Notifying authorities and notifying bodies”)
Chapter V (“General purpose AI models”)
Chapter VII (“Governance”)
Chapter XII (“Sanctions”) with the exception of Art. 101 (fines for GPAI providers).
Art. 78 (“Confidentiality”)
2026 – 02 February – “High-Risk” Defined
The Commission will publish guidelines specifying the practical implementation of Art. 6 together with a comprehensive list of practical examples of use cases of AI systems that are “high-risk” and “not high-risk”. This is from Art. 6(5).
2026 – 02 August – Most Obligations Apply
The majority of the AI Act’s provisions will apply to AI systems with transparency risks, new high-risk AI systems as defined in Annex III, and existing high-risk AI systems that are significantly modified. During this 24-month transitional period, accompanying measures like delegated acts, guidelines, and standards will be established to ensure standardized and legally compliant implementation and enforcement. From Art. 57(1) and Art. 113 this will be:
National AI regulatory sandboxes are established
The remainder of the AI Act (except Art. 6(1) and obligations delayed by grandfathering in Art. 111)
2027 – 02 August – Full Application
Provisions regarding GPAI models placed on the market before 02 August 2025, new high-risk AI systems, and high-risk AI systems listed in Annex I will become fully applicable. This includes AI systems that are products covered by EU harmonisation legislation and subject to third-party conformity assessments. From Art. 111(3) and 113(c), this will be:
Providers of general-purpose AI models that have been placed on the market before 02 Aug 2025
Article 6(1) effective and the corresponding obligations
2030 – 02 August – Final Grandfathered Obligations
The AI Act will apply to providers and deployers of high-risk AI systems (ii) intended to be used by public authorities (ii) subject to significant changes in their designs as from 02 August 2026, and (iii) placed on the market or put into service before 02 August 2026. This is from Art. 111(2).
2030 – 31 December – Final Grandfathered Obligations
The AI Act will apply to AI systems that are part of large-scale IT systems as specified in Annex X, and which were placed on the market or put into service before 02 August 2027. This is from Art. 111(1).