A STUDY OF THE EUROPEAN UNION’S LEGISLATION ON ARTIFICIAL INTELLIGENCE

Abstract

The European Union’s Artificial Intelligence Act is mostly about high risk artificial intelligence systems, and these are heavily regulated. It also sets out a narrower category for low risk artificial intelligence systems, with lighter legal obligations, mainly relating to transparency. This means developers and deployers have to ensure users are clearly told they are talking to AI, as with chatbots and deepfakes. In contrast, artificial intelligence systems with minimal risk are not regulated. Most of the AI applications already on the EU single market fall into this category, including AI enabled video games and spam filters. As for liability, the Act makes providers of high risk artificial intelligence systems primarily liable, no matter where they are based and also applies to providers outside the EU if their systems are used in the EU. Obligations of deployers are relatively lighter and apply to deployers within the EU and users from third countries operating within the EU. GPAI model providers are subject to a range of obligations, such as technical documentation and copyright related obligations. Where such models are made available under a free licence, providers must also publish summaries of training data and further safeguards apply where the model creates a systemic risk.