The AI Act enters into force

2024-08-01 · Publisher : Union européenne · Domain : Artificial Intelligence

Regulation (EU) 2024/1689 on artificial intelligence enters into force on August 1, 2024. It starts a staggered application timeline running to 2027, from bans to obligations for high-risk systems.

On August 1, 2024, Regulation (EU) 2024/1689 on artificial intelligence, known as the AI Act, entered into force, twenty days after its publication in the Official Journal of the European Union on July 12. Adopted by the European Parliament on March 13 and by the Council on May 21, it became the first comprehensive legal framework dedicated to AI. ### What happened - The text classifies AI systems into four risk levels, as presented by the European Commission: - minimal risk (spam filters, video games): no obligations, voluntary codes of conduct; - specific transparency risk (chatbots, generated content): inform users and label content; - high risk (medical software, recruitment and so on): risk management, data quality, clear information, human oversight; - unacceptable risk (social scoring, for example): banned. - An AI Office is set up within the Commission to oversee general-purpose AI models. ### Application timeline | Date | Applicable obligations | |---|---| | August 1, 2024 | Entry into force | | February 2, 2025 | Bans and AI literacy obligation (training) | | August 2, 2025 | General-purpose AI models, governance, penalties | | August 2, 2026 | General application, including most high-risk systems | | August 2, 2027 | High-risk systems embedded in regulated products | ### Key figures Fines can reach 35 M EUR or 7% of annual worldwide turnover for prohibited practices, with lower amounts for other breaches. ### Why it is a turning point For the first time, developing or deploying AI in the European Union is subject to legal obligations proportionate to risk. Like the GDPR for personal data, the text applies to non-European players as soon as they target the European market. ### Caveats Application depends on harmonized standards, guidelines and a code of practice for general-purpose AI, some of which were delivered late. In November 2025, the Commission proposed, in a digital simplification package, to adjust the timeline of some high-risk obligations: the dates above reflect the 2024 text. ### Key takeaways August 1, 2024 started the countdown: mapping AI uses, qualifying their risk level and training teams became obligations rather than just good practice.

Tags : IA, Régulation, AI Act, Europe, Conformité