On 7 December, EIF organised a debate hosted by Dragos Tudorache MEP on the proposed AI Act with a focus on AI Governance in Europe. The MEP was joined by the following guest speakers:
Kilian Gross, Head of Unit 'AI Policy Development and Coordination', DG CNECT, European Commission
Nichole Francis Reynolds, VP & Head of Global Government Relations, ServiceNow
Koen Cobbaert, Senior Manager – Quality, Standards and Regulations, Philips
Miguel Valle del Olmo, Deputy Director General for Artificial Intelligence and Digital Enabling Technologies, The Spanish Ministry of Economic Affairs and Digital Transformation
MEP Dragos Tudorache opened his speech by declaring that governance is essential to find a balance between the innovation-centric and the human-centric approach to AI. A European layer of responsibility has been added to the initial proposal, not only for monitoring and overseeing implementation, but also in terms of enforcement. He focused on sandboxes, a fundamental element that demonstrates the necessity of a European-level governance as well as the already-mentioned balance between innovation and human-centricity of the Regulation. Furthermore, stakeholders should be structurally included in the decision-making process. Since the AI Act is the first horizontal legislation on AI in the world, it is critical to strive for convergence and to be mindful of the approach that the US and other like-minded partners will adopt. In conclusion, MEP Tudorache underlined the key role of standardization: “the regulation needs to have standards that should be relevant globally”.
The European Commission representative, Kilian Gross, affirmed that, despite the numerous amendments, the structural concepts of the AI Act are shared: risk-based approach, horizontal legislation, use of standards and requirements for high-risk AI applications. A good balance between innovation and regulation is critical. He stressed that the proposal of the Commission gave additional powers to the Member States through a Board that will have to: help market surveillance authorities in case of conflicts; give recommendations on the delegated acts and the implementing acts; work on the guidance to assist the market; have powers for cross-border investigations and international matters. Mr. Gross concluded his speech by acknowledging the importance of having a regulatory sandbox as a protected area in which small companies would develop the system under some supervision.
Nichole Francis Reynolds highlighted the work that European institutions and the TTC are doing in terms of global harmonisation, which is fundamental for compliance purposes from both the industry and the consumer perspective. Artificial intelligence is a cutting edge technology that can strengthen the global economy, but has to be used responsibly. However, she recognised that AI requires coding, developing and providing. She stressed the role of evaluation, diving into the pivotal role of sandboxes: “It is important to allow industries like ours to test and use open source again responsibly so that we know what works, what the best practices are, what the best practices are not”. In conclusion, Ms. Francis Reynolds highlighted the function of global harmonisation to guarantee the benefits of AI for businesses, governments, consumers and the global economy.
Koen Cobbaert acknowledged that, in the medical device sector, trustworthy AI systems can improve health care and patient outcomes. However, a robust regulatory framework is needed for ensuring legal clarity within the AI Act and between the AI Act and sectorial legislation, such as the medical device regulation. Otherwise, legal uncertainty might impact patients adversely, and discourage investments in medical devices in the EU. In the current Medical Device Regulation, there are supply issues due to the difficulties of implementing the regulation in the EU. Despite the Council's improvements to the text, he acknowledged the big gap between policy makers, computer scientists and safety engineers. In conclusion, Mr. Cobbaert declared that the AI Act should provide access to the training, testing and validation of data sets.
Miguel Valle del Olmo, representing the Spanish Ministry of Economic Affairs and Digital Transformation, focused on three significant elements concerning AI governance: coordination processes leading to governance frameworks; regulatory experimentation; future steps on the implementation of the AI Act. He defined AI as a clear paradigm of horizontal technology. Coordination processes among different countries are pivotal and should include the definition of a concrete governance framework. According to him, the European Artificial Intelligence Board is an effective governance mechanism including national authorities of Member States and other actors, but also ensuring a proper harmonisation of administrative practices. In terms of regulatory experimentation, Mr. Valle del Olmo admitted that the AI Act itself inspired the design of the sandbox. The proposed sandbox would include a monitoring phase of high-risk systems with the aim to detect incidents or risks.