Liability for Users of Artificial Intelligence

Liability for Users of Artificial Intelligence

Anne Marie Frøseth

In this presentation, Professor Anne Marie Frøseth from the Faculty of Law in Bergen presents issues she has worked with in her research on liability for damages caused by artificial intelligence.

The presentation provides an overview of the basic features of the European Parliament's and the European Commission's proposals for harmonised liability rules for users of artificial intelligence from 20 October 2020 and 28 September 2022, respectively. The future of a supranational EU liability regime for users under EU tort law is highly uncertain. The proposed legislative text is complex and creates unclear boundaries with the EU product liability regime.  Some areas of overlap with the European Commission's proposal for revision of the Product Liability Directive from 28 September 2022 will therefore also be discussed.

The EU's proposals for harmonised rules in the area of tort law have been strongly criticised. Firstly, it can be questioned whether a separate user liability regime is the right approach, in light of the fundamental considerations and principles in EU’s regulatory strategy for artificial intelligence. Secondly, it can be questioned to what extent the assumption that the national tort rules are currently inadequate is correct, and , in regard to this specific issue fault-based liability and non-statutory strict liability in Norway will be used as examples. And last but not least, there are many indications that the European Commission's proposal in particular will have a very narrow scope of application. The limitations of the proposals will therefore also be discussed, and to some extent compared with the proposed revision of the product liability directive. 

The topic is relatively extensive, but if there is room for further discussion, it may be interesting to discuss a couple of models for enterprise liability (in a broad meaning) and the extent to which they are suitable for filling the "responsibility gap" that is claimed to exist as a result of the special risk profile of AI systems.

The presentation will be held in Norwegian.