As autonomous AI agents—software systems designed to pursue goals independently—become more prevalent, legal experts are grappling with the question of liability when these systems cause harm. According to Professor Jeannie Paterson of the University of Melbourne, the current legal consensus is that the individual or organization that deploys an AI agent remains responsible for its actions. This holds true even in cases where the harm was unintended, provided that the outcome was reasonably foreseeable.

However, the rise of "agentic" AI, which can operate with a high degree of independence, has introduced new challenges for legal frameworks. While the principle of human responsibility is clear in theory, experts point to a growing "murkiness" in practice. As these systems make decisions that their creators may not have explicitly programmed, determining the boundaries of foreseeability and negligence becomes increasingly difficult. The discussion has gained urgency following Australia's first documented case of an AI agentic "accident," which has highlighted the gap between traditional legal doctrines and the realities of modern automation. Legal scholars suggest that while existing laws provide a starting point, the unique nature of autonomous software may eventually require more specific regulatory guidance to address scenarios where human oversight is limited or indirect.