The European Chamber gives the green light to the AI Act, i.e. the establishment of European standards on Artificial Intelligence.
The approval of the text in plenary came by a large majority: 523 votes in favour, 46 against and 49 abstentions. The EU is the first in the world to adopt rules on artificial intelligence.
The regulation aims to protect fundamental rights, democracy, the rule of law and environmental sustainability from high-risk AI systems, while promoting innovation and ensuring Europe leads the field.
The legislation provides for obligations based on the possible risks and level of impact of AI systems.
In other words, the new rules outlaw some AI applications that threaten citizens' rights, including biometric categorization systems based on sensitive characteristics, emotion recognition systems in the workplace and in schools, social scoring , predictive policing practices.
However, there are exceptions.
In some specific situations expressly provided for by law, law enforcement agencies may use biometric identification systems, which are otherwise prohibited.
"Real-time" identification can only be used if strict guarantees are respected, for example if the use is limited in time and space and subject to judicial or administrative authorization.
Permitted uses include, for example, searching for a missing person or preventing a terrorist attack.
As for high-risk systems, i.e. those that could cause significant damage to health, safety, fundamental rights, the environment, democracy and the rule of law, there is an obligation to assess and reduce risks, maintain records of use, be transparent and accurate and ensure human surveillance.
One of the new features, compared to the Commission's proposal, is the provision of rules for AI systems for general purposes, which will have to satisfy certain transparency requirements and respect EU copyright rules during the training phases of the various models.
More powerful models, which may pose systemic risks, will have additional obligations, such as carrying out model evaluations, assessing and mitigating systemic risks, and reporting on incidents.
Furthermore, artificial or manipulated images and audio or video content, so-called deepfakes, must be clearly labeled as such (watermarking).
Finally, EU countries will have to establish and make accessible at national level regulatory experimentation spaces and testing mechanisms in real conditions (sandboxes), so that SMEs and start-ups can develop innovative AI systems and train them before introducing them on the market.
The regulation will enter into force twenty days after its publication in the Official Journal of the EU and will begin to apply 24 months after its entry into force, except with regard to prohibitions on prohibited practices, which will apply from six months after its entry into force. entry into force;
codes of good practice (nine months later);
rules on general purpose AI systems, including governance (12 months) and obligations for high-risk systems (36 months).
Reproduction reserved © Copyright ANSA