Prohibited artificial intelligence practices in the proposed EU artificial intelligence act (AIA)

Abstract

Artificial intelligence (AI) now forms a more and more important part of human lives. After years focussed on the development of AI, the initial hype about its many expected benefits has gradually given way to rising ethical concerns about its inherent risks and dangers. Efforts to confront and contain the most serious risks related to AI have now prompted a number of legislative or regulatory proposals at the national, regional and global level. One of the most comprehensive regulatory initiatives is the European Union's proposal for an Artificial Intelligence Act (AIA), which was released in April 2021 with a view towards establishing a legal framework for trustworthy AI. To this end, the draft AIA pursues a proportionate horizontal and risk-based regulatory approach to AI, broadly classifying AI into the categories of unacceptable risks, high risks and low or minimal risks. The unacceptable risks are those that are deemed to contravene European Union values, and therefore, they are considered to be ‘prohibited AI practices’ by Article 5 AIA. The prohibited AI practices are classified into four categories, namely 1) AI systems deploying subliminal techniques; 2) AI practices exploiting vulnerabilities; 3) social scoring systems; and 4) ‘real-time’ remote biometric identification systems. The proposed regulatory approach, however, appears problematic given the four categories’ inherent interrelatedness and the numerous possibilities for their mutual combination and entwinement. It is also problematic from the perspective of the human mind, as each of the four categories alone allows for the manipulation of human thought and behaviour, thereby endangering freedom of thought and other fundamental rights. In the context of the proposed AIA, both aspects give rise to unknown and unsolved conundrums that create difficult regulatory challenges that raise the necessity to also look at the wider implications of these technologies for the entire legal system. As these conundrums often find their expression in paradoxes and oxymora, this article calls for a wider interdisciplinary debate and advocates a different regulatory strategy using these concepts to transcend the limitations inherent in dualistic or dichotomous modes of legal thinking.