AI responsibility and the illusion of control: doctrines, gaps and the EU response

Purpose This paper aims to explore whether traditional liability doctrines can adequately address damages caused by the actions of artificial intelligence (AI) systems and whether AI systems themselves can be held accountable in cases of improper performance. In addition, the paper examines certain key aspects of the EU AI Act while considering what the law ought to be to fully accommodate the unique characteristics and consequences of AI systems. Design/methodology/approach This paper adopts a doctrinal legal analysis methodology, drawing on relevant legal sources, including legislation and academic literature. By identifying structural gaps, challenges and inadequacies in the current legal system, the analysis informs the need for legal reform. Findings Despite the increasing sophistication and self-learning capabilities of autonomous AI systems, current legal doctrines and legislative initiatives continue to conceptualise such systems as passive instruments. This persistent disconnect between technological advancement and existing legal frameworks underscores the need for early regulatory intervention during design and production, rather than waiting until technological outcomes have already emerged. It is also argued that an exclusive reliance on detailed documentation, risk assessments and compliance measures can be prohibitively costly, both economically and practically and may not lead to meaningful results. Originality/value This paper presents a thoughtful examination of the legal recognition of responsibility for the actions of AI systems, clarifying gaps in existing legal approaches that hinder their effective deployment. It provides insights to inform future research and policy development and offers recommendations for a more supportive legal framework that fosters both innovation and the effective allocation of liability in the AI context.