2026: New Game Rules for Artificial Intelligence in the EU
December 8, 2023: The Commission, the Parliament and the European Council reach an agreement on the final version of the AI act, concerning the regulation of artificial intelligence. Although the text has not yet been published, here are some key points that have reached consensus:
Make visible how AIs (artificial intelligences) have been trained, as well as their consequences on society. The content on which AIs are trained, as well as tests to anticipate potential deviations, will be detailed by the designers and examined before these solutions enter the European market. The more powerful and/or risky an AI is, the more it will be accompanied by experts (external to its designers) to evaluate it and develop mechanisms to limit its impact.
Use remote biometric identification systems: within a judicial authorization where offenses are strictly defined, cameras or drones may use facial recognition. The goal will be to search for an individual convicted or suspected of having committed a serious crime.
Prohibit any social scoring, as well as any use of AI exploiting the vulnerability of users, based on their religious, philosophical, political opinions, race, sexual orientation or emotions.
Prohibit the creation of facial recognition databases, extracted from images of individuals found randomly on the internet or through video surveillance. This applies both in the workplace and in educational institutions. In case of non-compliance with this principle, the individuals concerned can file a complaint and must receive meaningful explanations.
- Impose fines ranging from 1.5% to 7% of their turnover on convicted companies.
To implement this regulation and the resulting sanctions, a new European organization should be established in the coming years. This institution may, as a last resort, reject the entry of an AI service into European territory.
Effective in 2026, this regulation urges member countries to work on legislation that meets European requirements. However, this regulation will not apply to AI solutions used solely for military purposes, so as not to hinder member states in ensuring their national security.
Sources :
- European Parliament - Artificial Intelligence Act: deal on comprehensive rules for trustworthy AI
- Usine Digitale - L'Europe trouve un accord sur la régulation de l'intelligence artificielle
- France Culture - Le biais d'Aurélie Jean : "AI Act, une bonne nouvelle"
- La Revue du Digital - L’Union européenne décide d’un accord global sur la régulation de l’IA
[Photo copyright: European Union]
Support us by sharing the article: