The European Union’s Regulatory Approach to Artificial Intelligence and the Detection of Deepfakes

In April 2021, the European Commission published a proposal for a regulation on laying down harmonized rules on artificial intelligence (“EU AI law”). The regulatory approach selected by the Commission for this initiative goes far beyond the numerous papers published in recent years on law and ethics regarding the use of AI. This paper looks specifically at the regulation of deepfakes. Deepfakes are fake videos produced with artificial intelligence tools. They can be used for numerous purposes, legitimate or illegitimate, including intentional manipulation of political decision-making processes; for example during electoral campaigns. The proposed paper investigates possible contributions that an AI-based deepfake-detector could make in responding to disinformation, which is a threat to democracy. Disinformation is certainly not a new topic, but recent technological developments relating to AI-generated deepfakes have increased the manipulative potential of video and audio-based contents spreading online. As some uses of deepfakes show a potential for legitimate creative, educational, or political purposes the regulatory approach to be selected is necessarily complex and requires a better understanding of differentiated thresholds regarding their legality in various case-based applications. In this paper, we will first investigate the Commission’s regulatory approach with respect to deepfakes. In the second step, we will consider technological and legal strategies that can react to some of the challenges posed by deepfakes, in particular how a deepfake detector at the disposal of the general public could contribute to improving public digital literacy, in light of the increasing complexity of assessing whether online video or audio contents are authentic.