The Chamber of Commerce and the Chamber of Trades have issued a joint opinion on bill 8476, which aims to implement certain provisions of Regulation (EU) 2024/1689 on artificial intelligence (AI Act).

Whilst the two professional chambers welcome the government’s commitment to the rapid implementation of the AI Act, they stress the importance of creating an attractive ecosystem and make several observations and recommendations regarding regulation as well as the establishment of regulatory sandboxes.

Make the CNPD the “Luxembourg Data and AI Authority”

With regard to regulation, the bill gives the National Data Protection Commission (CNPD) the role of horizontal supervisory authority by default. It would be called upon to oversee seven sectoral authorities: the Autorité de contrôle judiciaire, the Financial Sector Supervisory Commission (CSSF), the Supervisory Authority for the Insurance Sector (Commissariat aux assurances, CAA), the Luxembourg Institute of Standardisation, Accreditation, Safety and Quality of Products and Services (Ilnas), the Luxembourg Regulatory Institute (ILR), the Agence luxembourgeoise des médicaments et produits de santé (ALMPS) as well as the Autorité luxembourgeoise indépendante de l’audiovisuel (Alia).

The architecture has given rise to reservations on the part of both chambers, which believe that it runs counter to “clear, simple and accessible governance.” They argue in favour of the adoption of common rules of procedure and the creation of working groups bringing together the various market surveillance authorities.

In their view, this would make it possible to limit the risks of divergent--or even contradictory--practices, procedures or decisions between the competent authorities. They also consider it essential to make a clear distinction, within the CNPD, between the tasks relating to data protection (GDPR) and those arising from the AIAct. To this end, they suggest renaming the CNPD the “Luxembourg Data and AI Authority.”

Making economic arguments out of sandboxes

With regard to the introduction of regulatory sandboxes--test environments designed to facilitate the development and validation of AI systems before they are put on the market--the two chambers want to make this an argument for the country’s attractiveness. This could be “a strategic niche for Luxembourg, positioning it internationally as an entry point for companies from third countries that want to set up in the EU and test the compliance of their AI products.”

How? They propose key features for the sandboxes set up by the CNPD, including low-cost participation, close accompaniment, short duration of tests to speed up time to market, high guarantees of confidentiality, proactive and competent specialists, a close link between the exit report and the compliance procedure, and clear national rules on the liability regime.

They also recommend fostering a wider ecosystem of interoperable sandboxes, by exploring the possibilities offered by the AI Act: real-life trials with the Meluxina supercomputer, collaborations with other member states, or even setting up a specific sandbox for the financial sector, piloted by the CSSF.

This article was originally published in .