Contractual Liability for the Use of AI under Dutch Law and EU Legislative Proposals
de Graaf, I.S. Wuisman · Information technology and law series/Information technology & law series · 2022
In this chapter, the contractual liabilityLiability of a company (the ‘user’) using an AI system to perform its contractual obligations is analysed from a Dutch law and EU law perspective. In particular, we discuss three defences which, in the event of a breach, the user can put forward against the attributionAttribution of that breach to such user and which relate to the characteristics of AI systems, especially their capacity for autonomous activity and self-learning: (1) the AI system was state-of-the-artState-of-the-art when deployed, (2) the user had no controlControl over the AI system, and (3) an AI system is not a tangibleTangible object and its use in the performance of contractual obligations can thus not give rise to strict liabilityLiability under Article 6:77 of the Dutch Civil Code. Following a classical legal analysis of these defences under Dutch law and in light of EU legislative proposals, the following conclusions are reached. Firstly, the user is strictly liable, subject to an exception based on unreasonableness, if the AI system was unsuitable for the purpose for which it was deployed as at the time of deployment. Advancements in scientific knowledge play no role in determining suitability. Secondly, a legislative proposal by the European ParliamentEuropean Parliament liability proposal allows the user to escape liabilityLiability for damage caused by a non-high-risk AI system if the user took due care with respect to the selection, monitoring and maintenance of that system. Thirdly, the defence that the user is not liable because an AI system is not a tangibleTangible object is unlikely to hold.