The Need for a Legal Standard of Care in the AI Environment

Abstract

Artificial intelligence (AI) has risen to legal debate over legal liability in-volved in an incident. An intelligent machine can learn through experience and adapts its decisions accordingly. As such, if an intelligent machine’s be-haviour causes harm, the developer and the machine's owner may argue that the autonomous nature of AI systems has broken the causal link. The diffi-culty of determining who is liable for a harmful behaviour of an AI system is accentuated by the fact that tracing back the decision-making process of an AI system is not always possible. This paper aims to put forward a definition of a duty of care for developers and users of AI systems that could be the basis for the investigation of liability while seeking predictability of the allo-cation of legal liability in many cases involving AI incidents. The paper ex-amines some guidelines on ethics for AI to discern essential elements of the duty of care in the AI environment. The paper argues that a uniform minimum standard of care should be adopted internationally through model laws or even an international convention. A uniform standard of care should be enforced by State control rather than self-regulation by the AI industry. A licensing or certification requirement for AI products should be implemented to verify that the elements of the duty of care have been satisfied to control AI production and import/export relations. Violation of the standard of care can be an objective ground to negate or allocate negligence, especially when verifying errors in the design of the relevant software or if explaining the AI system's behaviour is not possible. A clear standard of care would, this paper assumes, help promote AI development and use and would not create imped-iments to investment in AI production.

Keywords

• artificial intelligence
• ai regulation
• legal liability
• duty of care.