Liability for AI as for an Assistant or a Trainee? Old Rules for New Technologies

Abstract

The dynamic development of artificial intelligence extends into the legal services sector. One of the key questions arising in this context concerns the identification and allocation of liability for damage caused by the improper use of artificial intelligence (as illustrated by widely publicized cases where lawyers submitted pleadings to courts citing fabricated case law). Traditional private law institutes, such as liability for an assistant, may provide a general starting point for determining the person responsible for the use of a particular AI system. This article examines the issue of legal liability for damage resulting from the use of generative artificial intelligence, in particular large language models (LLMs), with a specific focus on the provision of legal services. The basis of our analysis lies in the analogy with traditional models of liability, in particular the principal’s liability for an assistant (under the Civil and Commercial Codes) and the lawyer’s liability for a trainee, both firmly embedded in Slovak law. Our aim was to explore whether these existing legal mechanisms sufficiently cover situations in which a lawyer relies on AI in their practice (for instance, by unjustifiably relying on a particular output without its due verification), or whether modifications or new rules should be considered. We tend, however, to conclude that the current normative framework does not require fundamental changes. The article also addresses respective regulation adopted by Slovak Bar association.