EU AI Act Enters into Force: Key Compliance Dates for Stakeholders
The long-anticipated AI Act was published in the Official Journal of the European Union on 12 July 2024 and will enter into force on 1 August 2024.
Laying down harmonised rules on the regulation of artificial intelligence across all EU Member States, there are important compliance deadlines of which all stakeholders should be aware:
General Applicability after 24 Months – 2 August 2026
The AI Act provides for a 24-month transitional period, to allow time for delegated legislation, guidelines, and standards to be drafted and published across all Member States. Most general provisions of the Act will apply from 2 August 2026, with some exceptions outlined below.
Prohibited Practices Provisions – Applicability after 6 Months – 2 February 2025
Article 5 of the AI Act outlines ‘prohibited practices’ which become unlawful on 2 February 2025. These include exploitative or manipulative AI tools that alter human behaviour in a harmful way, social scoring techniques, the creation of facial recognition databases through untargeted scraping of images online or from CCTV, emotional recognition AI tools, as well as the use of biometric identification systems based on highly sensitive characteristics, among others.
General Purpose AI Models, Governance and Sanctions Provisions – Applicability after 12 Months – 2 August 2025
2 August 2025 marks the applicability of Chapter V relating to general purpose AI models, with obligations around classification, and procedure, as well as additional obligations placed on providers of general-purpose AI models which fall under a ‘systemic risk’ category.
Exception - Applicability for Certain High-Risk AI Systems after 36 Months – 2 August 2027
An exception to the general date of application applies to certain high-risk AI products subject to harmonising EU legislation listed in Annex I of the Act, or AI systems that are a safety component of such products that require a third-party conformity assessment in advance of being placed on the market. AI systems that fall within this category have an additional 12 months from the date of general applicability, with full compliance required by 2 August 2027.
Grandfathering Provisions - Differing Timelines
There are a number of grandfathering provisions under the AI Act. Those of most general application relate to “high risk” AI systems in use prior to 2 August 2026. Such “high-risk” AI systems are not subject to the AI Act unless they are subject to “significant changes in their design”, or if they are used by public authorities (in which case they must comply by 2 August 2030).
Conclusion
The risk-based approach of the AI Act brings with it a complex set of deadlines, obligations and provisions which have significant consequences in the event of non-compliance. It is expected and hoped that guidance will be issued in the coming months for stakeholders.
If you require assistance with compliance under the AI Act, please do not hesitate to get in touch with one of our key contacts below, or your usual McCann FitzGerald contact.
This document has been prepared by McCann FitzGerald LLP for general guidance only and should not be regarded as a substitute for professional advice. Such advice should always be taken before acting on any of the matters discussed.
Select how you would like to share using the options below