Ladies and Gentlemen,

It is my distinct pleasure to give some views on the implementation of the AI Act as the chair of the European Working Group of Competent Authorities on AI. We find ourselves at the nexus of a technological revolution, where artificial intelligence has ceased to be a distant future and has become an integral part of our present.

Three years ago, before the AI Act was even proposed, we recognized the imperative need to supervise AI systems. That’s when we started with the European Working Group of Competent Authorities on AI. This wasn't a forward-looking aspiration; it was a practical necessity. AI systems do not exist in a vacuum. They are part and parcel of the fabric of our society, touching upon various sectors and legislations, be it the Medical Device Regulation (MDR) or general legislation such as the GDPR or NIS-D. Our role as Competent Authorities is not confined to the supervision of the AI Act alone; our scope is far broader and more intricate. We need to keep that in mind when implementing the AI Act.

The AI landscape is rich and varied across our Member States. Each country brings to the table its unique perspective and methodology. Consider the distinct approaches of Spain and the Netherlands—Spain has chosen more the path of a singular national competent authority, while the Netherlands chooses a more decentralized approach, with a main role for sectoral competent authorities. When we consider AI in medical devices, it is not sufficient to evaluate the AI system in isolation. It must be assessed within the medical context it is designed to operate within. A siloed view is no longer sufficient; a multidisciplinary approach is necessary for true comprehension and effective supervision. 
We need to keep an open mind and learn from each other to find the best way to supervise AI and the AI Act.

This diversity extends to our regulatory sandboxes. The Spanish model is one of structured conformity, a legalistic framework designed to ensure adherence to the rules. The Dutch model, conversely, embraces a pragmatic and accessible strategy, creating an environment where an AI service center and a regulatory sandbox coexist to foster innovation while promoting the adoption of the AI Act. We need to be aware of the different options there are for implementing parts of the AI Act in the Member States and see what the impact will be.

Legislation, however powerful on paper, is only as effective as its implementation and enforcement in reality. Much of our discussion has focused on the requirements for AI systems, their developers, and their producers. But what of our ability, as competent authorities, to supervise under the AI Act? This remains a pivotal question. Our collaboration with UNESCO aims to establish not just frameworks but effective, actionable methods of supervision.

I am looking forward to an effective and fruitful discussion during the conference.

Huub Janssen

Chair of the European Working Group of Competent Authorities on AI 

Chair of the Dutch Working Group of Competent Authorities on AI 

Responsible for supervising AI at the Dutch Authority for the Digital Infrastructure