Abstract. The rapid deployment of artificial intelligence systems in consequential domains — including criminal justice, employment screening, credit allocation, healthcare, and social benefit determination — has generated urgent scholarly and public debate about the ethical foundations of algorithmic decision-making. This paper provides a comprehensive interdisciplinary review of the principal ethical challenges posed by AI systems in high-stakes contexts, with particular emphasis on the problem of algorithmic bias and its mechanisms, manifestations, and remediation. Drawing on scholarship from computer science, philosophy, law, sociology, and political science published between 2016 and 2025, the review examines how structural biases embedded in training data, model design choices, and institutional deployment contexts translate into discriminatory outcomes for marginalized populations. The paper further analyzes existing and proposed governance frameworks — including technical fairness interventions, regulatory approaches, and participatory design methodologies — and evaluates their adequacy in addressing the multidimensional nature of AI ethics. The central argument is that algorithmic bias is not primarily a technical problem admitting a technical solution, but rather a sociotechnical phenomenon that requires comprehensive institutional, regulatory, and democratic responses.