Quote this article

Barichella, A. « Regulating artificial intelligence at the EU level: obstacles and prospects», Policy Brief, Institut Jacques Delors, November 2023

This paper aims to examine the obstacles and prospects for regulating artificial intelligence at the EU level. Firstly, a number of obstacles remain, including the difficulty of defining AI and the appropriate regulatory scope, the continued sway exercised by lobbying groups, along with the velocity of change in the AI industry which makes it challenging for regulators to keep-up. Secondly, in terms of the AI Act itself, the EU has chosen an approach that exhibits many strengths, relying on a ‘technology-neutral’ definition and setting out a ‘riskbased’ approach whereby AI systems are regulated according to the degree of risk they pose to society (the four categories include ‘unacceptable risk’, ‘high-risk’, ‘limited-risk’ and ‘low or minimal risk’). The legislation still suffers from a number of inadequacies, however. These include insufficient flexibility in adapting to the speed of evolution in this sector, an over-emphasis on individual risks and thus weaker consideration of the broader societal-level impacts, and inadequate compliance frameworks which often rely on self-assessment. Thirdly, the EU is well positioned as a ‘first mover’ in the field of AI regulation to play a key role in influencing both national rules and international standards. Due to the complex and multi-faceted nature of AI technologies, the EU should consider a model such as the international regime for the prohibition of chemical weapons, with an additional forecasting unit, to establish global rules and monitoring of AI.