The European Parliament adopts the AI Act
2024-03-13 · Publisher : Parlement européen · Domain : Artificial Intelligence
Meeting in Strasbourg, MEPs adopt the Artificial Intelligence Act by 523 votes to 46, with 49 abstentions: the world's first binding law dedicated to AI clears its decisive stage.
What changes
- A landslide vote: 523 in favor, 46 against and 49 abstentions in plenary.
- Risk-based approach: the more dangerous an AI use, the heavier the obligations.
- Clear bans: social scoring, emotion recognition at work and in schools, predictive policing based solely on profiling.
- General-purpose AI models receive their own transparency obligations.
- Staggered application: 6 months for bans, 12 months for general-purpose AI, 24 to 36 months for the rest.
On March 13, 2024, the European Parliament adopted the Artificial Intelligence Act (AI Act) in plenary. Proposed by the Commission in April 2021 and negotiated with the Council until the political agreement of December 2023, the text became the world's first binding legislation dedicated to AI. ### What happened - The text was adopted by 523 votes in favor, 46 against and 49 abstentions. - It classifies AI systems by level of risk and bans practices deemed contrary to fundamental rights: untargeted scraping of facial images to build recognition databases, biometric categorization based on sensitive characteristics, emotion recognition in the workplace and in education, social scoring, predictive policing based solely on profiling, and behavioral manipulation. - High-risk systems (health, recruitment, critical infrastructure and so on) will have to assess and reduce risks, keep logs and ensure human oversight. - General-purpose AI models will have to meet transparency requirements, including compliance with copyright law and publishing a summary of training data. ### Timeline set out after the vote | Step | Delay after entry into force | |---|---| | Bans | 6 months | | Codes of practice | 9 months | | General-purpose AI | 12 months | | General application | 24 months | | High-risk systems embedded in products | 36 months | The text still had to undergo legal and linguistic review, then be approved by the Council, which happened on May 21, 2024. ### Why it is a turning point The European Union set the first comprehensive regulatory framework for AI, with extraterritorial reach: any provider placing a system on the European market is covered. Co-rapporteur Brando Benifei summed it up: “We have finally the first binding law worldwide on artificial intelligence.” ### Caveats The most concrete obligations depended on implementing acts, harmonized standards and codes of practice still to be drafted at the time of the vote. Part of the ecosystem feared a compliance burden that would penalize European players. ### Key takeaways The March 13 vote set the framework; entry into force followed on August 1, 2024. Companies developing or deploying AI in Europe then needed to map their uses by risk level.
Links
Tags : IA, Régulation, AI Act, Europe