ARTICLE
22 December 2023

EU Regulation On Artificial Intelligence Is Coming

EH
ELVINGER HOSS PRUSSEN, société anonyme

Contributor

Independent in structure and spirit, Elvinger Hoss Prussen guides clients on their most critical Luxembourg legal matters. Committed to excellence and creativity in legal practice, our firm delivers the best possible advice for businesses, institutions and entrepreneurs, playing a unique role in the development of Luxembourg as a financial centre.
After two years of negotiation and a final session of trilogue that lasted hours, on 9 December 2023 the European Parliament and the Council of the EU finally reached a political agreement on the coming...
European Union Technology
To print this article, all you need is to be registered or login on Mondaq.com.

After two years of negotiation and a final session of trilogue that lasted hours, on 9 December 2023 the European Parliament and the Council of the EU finally reached a political agreement on the coming regulation on artificial intelligence ("AI Act"). Ursula von der Leyen, President of the European Commission, welcomed this political agreement, pointing out that the AI Act "is the first-ever comprehensive legal framework on Artificial Intelligence worldwide".

The final text, however, is not yet available since work must continue at technical level, and the drafting of the AI Act - based on the compromise reached - has yet to be finalised. The European Parliament and the Council of the EU will in principle formally adopt the AI Act in the next weeks.

The AI Act concerns AI systems that will be placed on the EU market and used in the EU. The AI Act aims at striking a balance between the protection of fundamental rights and EU values, and the stimulation of innovation and investment in the field of artificial intelligence.

During the trilogue sessions, points of disagreement arose between Member States, in particular with respect to the inclusion of foundation models within the scope of the regulation and the use of real-time remote biometric identification for law enforcement purposes. Foundation models are also called "general purpose AI models" and are systems capable of performing a wide range of tasks (such as the generative AI model, GPT 4).

The AI Act adopts a risk-based approach, classifying AI systems from minimal or no-risk, to unacceptable risk. Minimal-risk AI systems will be subject to light transparency obligations, high-risk systems to stricter requirements and AI systems posing an unacceptable risk simply prohibited.

Designed on the same pattern as the GDPR, the AI Act should provide national surveillance authorities with the power to supervise the implementation of the AI Act at national level, while a European body would ensure coordination at European level. In the case of any infringement of the AI Act, the fines will be set as a percentage of the offending company's global annual turnover in the previous financial year or a predetermined amount, whichever is higher. Reduced caps should apply to SMEs and start-ups.

All the obligations of the AI Act will apply 24 months after its entry into force, with some exceptions. For example, prohibitions are supposed to apply 6 months after the entry into force of the AI Act.

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.

See More Popular Content From

Mondaq uses cookies on this website. By using our website you agree to our use of cookies as set out in our Privacy Policy.

Learn More