Jusletter IT

In 2021 the European Commission proposed the world’s first legal framework for Artificial Intelligence. The proposal follows a risk-based approach to AI-regulation, meaning that artificial intelligence systems will have to adhere to different requirements proportional to the risk they pose. Insofar as a system will be considered “high-risk”, the artificial intelligence will have to be compliant with the requirements on human oversight set out in Article 14 of the regulation. According to article 14 par. 4 lit (b) these requirements include the obligation to be aware of a so-called “automation bias”, which will be the focal point of the following analysis. The paper aims to give a short overview over relevant parts of the new proposal, clarify what constitutes an “automation bias”, highlight the difficult compliance with this seemingly small requirement and propose possible solutions.