Following the progress report of the Slovenian presidency, circulated at the end of November, The Guild voices its concerns on the burdens the European Commission’s proposal for an Artificial Intelligence (AI) Act may impose on research and researchers and calls for the Commission to involve researchers in the future monitoring and amendment processes of the Act.
The European Commission published its proposal for a regulation laying down harmonised rules on Artificial Intelligence (AI Act) in April 2021. Since then, the Working Party on Telecommunications and Information Society in the Council of the EU has been advancing its examination of the proposal. Key points remain nevertheless to be discussed during the French presidency of the Council and probably the following ones. A few days ago, the European Parliament reached an internal agreement that the committees on Internal Market and Consumer Protection and on Civil Liberties, Justice and Home Affairs will lead the internal discussions on the European Commission’s proposal. In consequence, negotiations in the European Parliament are likely to kick off shortly.
In this context, The Guild highlights the crucial importance of considering the perspectives of universities and other research actors in these negotiations, which may otherwise be dominated by categories of actors with their own vested interests and overlook crucial aspects for the future scientific and industrial leadership of Europe in AI.
The overall approach in the Commission’s proposal for an AI Act presents serious limitations. A broad concern is that the AI Act’s approach may be viewed as an attempt to regulate AI systems, because of their potential uses, instead of banning a set of practices. While recognizing how complex it is to define AI, The Guild also contends that the definition in the Commission’s proposal is so broad that it could be interpreted as encompassing non-AI technologies. Another challenge is to ensure that this definition is future-proof: as AI technologies are continuously developing at a fast pace, the Act and its annexes might become outdated quickly and stifle innovation.
Furthermore, there are concerns regarding the additional burdens the AI Act could impose on researchers applying for EU funding. As all EU grant proposals involving use or development of AI must already provide minimum information on potential ethic risks, The Guild emphasises that the AI Act should avoid increasing the burdens on researchers by requiring them to demonstrate that their research complies with the Act.
Finally, the European Commission should draw lessons from the implementation of the General Data Protection Regulation (GDPR), which, because of unharmonized interpretation and implementation across EU and EEA member states, has created legal uncertainties and obstacles to the exchange and reuse of (personal) data even for the purpose of research.
To address the possible concerns, The Guild proposes the following recommendations:
- Establish a high-level expert group, composed of academic researchers among others, which would advise on whether any technological progress requires an amendment to the annex or to the body text of the AI Act
- Set up regulatory frameworks that explicitly ban jeopardizing practices, such as social scoring and exploitation of the vulnerabilities of a specific group of persons, instead of banning AI technologies which may be used for these practices
- Give the European Artificial Intelligence Board the mandate to ensure the harmonised implementation of the AI Act in the member states
- Ensure that the AI Act does not result in the blanket prohibition of some AI systems, which would prevent their use even for the purpose of research activities
- Ensure that universities can act as ‘regulatory sandboxes’ and test AI systems in safe environments, and support research projects aimed at elucidating the concepts introduced in the AI Act and finding how AI developers can concretely comply with the Act
Jan Palmowski, Secretary-General of The Guild of European Research-Intensive Universities, commented: “The AI Act is a cornerstone of EU regulation not only in an area of crucial strategic and commercial importance. It also relates to a central arena of scientific research and research-led innovation. If the EU wants to lead a global transformation in Artificial Intelligence, it must respect and support the needs of researchers.”
This article was first published on 14 December by The Guild.