The To the Acts was officially approved by the European Parliament, thus becoming the world's first law on artificial intelligence. The legislative process ended with 523 votes in favour, 46 against and 49 abstentions. The law will come into force in two years.
The European Union thus becomes the first in the world to adopt rules to regulate artificial intelligence.
“I welcome the European Parliament's extraordinary support for our AI Act, the world's first comprehensive and binding rules for trustworthy AI. Europe is now a global regulator of standards in the field of artificial intelligence. We are regulating as little as possible, but as much as necessary!", the EU internal market commissioner commented on X Thierry breton.
What is the Ai Act
The The Ai Act represents a pioneering regulation which will apply uniformly across all EU Member States, consisting of 85 articles and attachments. It is the first legislative text that aims to comprehensively regulate the field of artificial intelligence. Originally proposed in 2021, it has been enriched with further amendments over time. The main purpose of the regulation is to ensure that the technology complies with Union rights and values, including human oversight, security, privacy, transparency, non-discrimination and social and environmental well-being.
A key feature of the Ai Act is the classification of applications of artificial intelligence based on the associated risk levels. The European Commission has identified four levels of risk:
- Unacceptable risk: includes systems that violate EU fundamental values and will not be permitted within EU borders.
- High risk: concerns systems with a controversial impact on the safety and rights of people. Although it is not prohibited to use, certain requirements are required to ensure safety and protection of rights.
- Limited Risk: Includes systems that do not pose a significant threat and must meet a limited set of requirements, including transparency in the use of AI.
- Minimal risk: There are no specific legal obligations in this category.
Ai Act: what is forbidden
In Europe, different categories of AI-based systems will be subject to various restrictions, up to being completely banned.
The AI Act prohibits the use of artificial intelligence for iremote biometric identification in real time and a posteriori in public spaces, as well as categorization based on sensitive characteristics. Also prohibited are predictive systems used by the police, i social scoring used by governments to analyze citizen behavior e emotional recognition systems in work and school environments. Furthermore, artificial intelligence systems used to influence elections and recommendation systems of social media platforms were added to the high risk list. These measures are aimed at protecting the human rights and privacy of citizens.
Generative artificial intelligence applications, such as ChatGPT or Google Gemini, fall into the high-risk category. The companies that develop them will have to provide detailed information on how their language models are trained, to comply with European copyright laws, and will have to clearly indicate when images, videos and audio have been generated by artificial intelligence, especially when they portray real people.
Other applications have been judged to be of limited risk and will have to meet limited requirements, mainly regarding transparency in the use of artificial intelligence. Finally, minimal risk applications are not subject to specific legal obligations.
Ai Act: heavy fines
The regulation provides for severe sanctions for those who violate it, which may arrive up to 35 million euros or 7% of global revenues of the responsible company.
Ai Act: the latest news
Compared to the Commission's proposal, the regulation introduces rules for general purpose artificial intelligence (AI) systems, which must be transparent and comply with EU laws on Copyright during model training. More powerful models, with possible systemic risks, must follow additional obligations such as risk assessments and mitigations and incident reporting. THE deepfake must be clearly labelled. EU Member States must establish regulatory testing spaces and real-world testing (sandboxes) to facilitate the development and training of innovative AI systems by SMEs and start-ups.
Aic Act: in force in two years
The regulation will become effective 20 days after its publication in the Official Journal of the European Union and will be applied 24 months after the date of entry into force. However, bans on prohibited practices will apply six months after entry into force, codes of practice nine months after, rules on artificial intelligence (AI) systems for general purposes, including governance, will apply after 12 months , while the obligations for high-risk systems will be applied after 36 months.
Last update 15,40am
