Artificial Intelligence and Legal Accountability: Frameworks for Algorithmic Governance

Abstract

The accelerating deployment of artificial intelligence systems in consequential decision-making credit scoring, employment screening, welfare eligibility, criminal justice, and healthcare has outpaced the development of legal accountability frameworks. This paper offers a comparative analysis of emerging regulatory responses to algorithmic harm across three jurisdictions: the European Union's Artificial Intelligence Act 2024, the United States' sectoral and state-law approach, and India's Digital Personal Data Protection Act 2023 and proposed Digital India Act. The analysis constructs a typology of algorithmic harms discrimination, opacity, safety failures, and informational manipulation and evaluates the regulatory tools addressing each type: risk classification, conformity assessment, data-protection impact assessments, transparency obligations, human oversight, and ex post liability. The paper argues that the EU's risk-based horizontal approach and the US sectoral approach each capture different but partial dimensions of algorithmic accountability, while the Indian framework remains under-specified. It proposes a convergent framework combining horizontal transparency obligations with sectoral risk regulation and a sustainable liability regime.