Abstract

The proliferation of Artificial Intelligence (AI) in decision-making across critical sectors such as healthcare, criminal justice, finance, and employment has raised pressing questions regarding legal liability and accountability. This article explores the multifaceted challenge of determining who is responsible when algorithmic decisions lead to harm or injustice. It begins by examining the structure and functioning of AI systems, particularly machine learning models, and identifies how the "black box" nature of these systems complicates legal scrutiny. Through real-world case studies, including Amazon’s biased hiring algorithm and the use of COMPAS in criminal sentencing, the article illustrates the tangible consequences of unregulated AI. It critically evaluates emerging legal responses, such as the EU’s proposed AI Act and suggestions for AI personhood, and considers alternative models like assigned liability, mandatory insurance, and human-in-the-loop oversight. The article argues that a coherent and proactive legal framework tailored to AI’s unique characteristics is necessary to ensure accountability, fairness, and redress. Furthermore, it advocates for embedding normative legal principles into AI governance and stresses the importance of international harmonization to prevent regulatory arbitrage. The overarching conclusion is that legal systems must evolve in tandem with technological innovation to safeguard human rights and societal trust. The liability question is not merely a legal dilemma but a fundamental test of democratic institutions in the digital age.

Keywords