2021: A Face Odyssey : An analysis of the proposed AI Act and its effect on current law and the police’s ability to use facial recognition technology

Sammanfattning: Artificial intelligence is becoming an increasingly important part of our lives and can be found in everything from fridges to phones. One of the applications of AI is the police use of facial recognition technology for law enforcement purposes. However, the use poses a major risk to fundamental rights. As part of the European Commission's initiative to create a Union fit for the digital age, the proposal for an AI Act was introduced in 2021 with the aim of setting the limit of permissible use of AI. In the act, the use of real-time facial recognition is prohibited except for a few exceptions which the police in the Union argue will severely hinder their work. However, the scope of the prohibition is ambiguous, and the act is riddled with various problems in its regulation of facial recognition used for law enforcement purposes. This thesis will therefore critically analyse the AI Act on the basis of three research questions. The first question examines how the AI Act will affect current law and the police’s ability to use facial recognition for law enforcement purposes. This thesis finds regarding real-time facial recognition, the act will replace LED as applicable law. However, for high-risk applications of FRT, there will be an interplay between the two regulations. When it comes to the effect on the police’s use of FRT for law enforcement purposes, this thesis finds that the police have no bigger reason for worry. The exceptions make a myriad of otherwise prohibited uses of real-time FRT permissible, and the use of post-FRT is not regulated in the act. The second question analyses the problems with the act and the thesis finds that there are essentially four major problems with the act. The exceptions allow for a disproportionate amount of otherwise prohibited uses of FRT, it is too difficult to interpret and the mechanisms for futureproofing are lacking. Furthermore, the interplay with Prüm II must be considered to a greater extent. The third and final question provides three different solutions to the problems. The first solution is to reduce the scope of the exceptions. The second solution is to revise the high-risk provision to make it easier to add new systems. Finally, this thesis finds that the introduction of a separate regulation for law enforcement use, like GDPR & LED, would be a good solution.