Artificial Intelligence Act: the EU’s first ever law on AI is officially published
On 12th July, the Artificial Intelligence Act (AI Act) has been published in the EU Official Journal (Regulation (EU) 2024/1689). Officially signed by the EU’s co-legislators on 13th June, the AI Act corresponds to the first binding worldwide horizontal regulation on artificial intelligence. It sets a common framework for the use and supply of AI systems in the EU while fostering the development and uptake of safe and trustworthy AI systems by both private and public actors.
The AI Act implementation timeline updated with final dates and relevant transition periodsSource: CECE
The new regulation establishes a classification for AI systems with different requirements and obligations tailored on a risk-based approach (the higher is the risk, the stricter the rules).
The AI act addresses the risks associated with specific uses of AI, categorising them into four levels of risk. As such, the AI systems with 'unacceptable' risks are prohibited. A wide range of 'high-risk' AI systems are allowed but subject to a set of requirements and obligations. AI systems presenting limited risks (‘transparency risk’) will be subject to information and transparency requirements, while AI systems presenting ‘minimal risk’ will not be subject to further obligations. The regulation also provides specific rules for general purpose AI (GPAI) models and lays down more stringent requirements for GPAI models posing a systemic risk with a significant impact on the internal market.
The first ever law on AI will enter into force on 1 August 2024, which corresponds to the twentieth day after its publication. This important step marks the starting date for the 24-month transition period before the new regulation becomes mandatory in all its parts. The AI Act application date is indeed set for 2nd August 2026. However, a gradual phasing-in of the AI Act key provisions is foreseen between today and 2026, with different transition periods:
- Six months after the entry into force (from 2nd February 2025), the general provisions (Chapter I) and prohibited AI practices rules (Chapter II) will start applying;
- Twelve months after the entry into force (from 2nd August 2025), the provisions on notified bodies (Chapter III Section 4), GPAI models (Chapter V), and governance structure (Chapter VII) will start applying. In addition, penalties (Chapter XII) with the exception of fines for GPAI models providers (Article 101) will also become applicable;
- Thirty-six months after the entry into force (from 2nd August 2027), Article 6(1) on high-risk AI categorisation for Annex I and corresponding obligations will start applying.
In the meantime, the European Commission will be working on the development of guidelines on the practical implementation of the AI Act, including the interplay of the regulation with the EU harmonisation laws listed in Annex I (e.g. Machinery Regulation), the application of the definition of ‘AI system’ and the term of ‘substantial modification’.
CECE will monitor the works of the European Commission with a view to support members in the interpretation of the main provisions and implementation of the new rules, especially when the application of the AI Act may have implications on the interpretation of the Machinery Regulation provisions.
More news