Enterprise Liability for Human-AI-Decisions

A Multidisciplinary Approach for Identifying Principals’ Duties of Care

Abstract

Although AI systems are constantly improving, in most areas, it seems unlikely that they will completely replace humans in the near future. Rather, AI systems and humans will increasingly work side by side. Integration of Human-AI collaboration holds great promise for many enterprises. Introducing AI while keeping humans “in the loop” may allow for the full utilisation of both AI and human potential. However, such collaboration also creates specific risks of damages, which raises the question of enterprise liability. In jurisdictions where the liability of a principal requires fault, as  generally in Germany, the answer to this question largely depends on the scope of the principal’s duties of care. These duties are examined in this paper. In the context of Human-AI-Decisions, the complexity of human decision-making, AI systems and, in particular, Human-AI cooperation can make it difficult to identify a principal’s duties of care. This paper shows how a multidisciplinary approach can help to overcome some obstacles. It does not provide an exhaustive list of the principal’s duties of care. Rather, it sets out a methodological framework and illustrates the value of non-legal AI research with specific examples. The paper also briefly discusses the proposed, but now abandoned AI Liability Directive.