GENERAL AI LEGAL PRACTICE HALLUCINATIONS PROFESSIONAL RESPONSIBILTY AND THE NECESSITY OF HUMAN OVERSIGHT

This paper examines the growing use of Generative Artificial Intelligence (AI) in legal practice, with particular focus on AI hallucinations, professional responsibility, human oversight, and legal judgment. It explores how AI tools are increasingly being used for legal research, drafting, document review, summarisation, and other professional activities, while also examining the risks associated with inaccurate or fabricated legal information. The study adopts a narrative review and doctrinal legal research methodology, examining academic literature, professional ethics guidance, regulatory developments, and selected judicial decisions. Particular attention is given to Mata v. Avianca, Inc., ABA Formal Opinion 512 (2024), the European Union Artificial Intelligence Act, and relevant developments in India's data-protection framework. Based on this analysis, the paper proposes a Human-Verified Legal AI Framework consisting of five stages: AI-assisted generation, verification of legal sources, human legal review, risk and ethical assessment, and final human approval. The framework adopts a risk-sensitive approach in which the degree of human oversight corresponds to the potential consequences of AI-assisted errors. The paper argues that responsible adoption of generative AI in legal practice should follow a human-led, AI-assisted model, where technology enhances professional capability while responsibility, accountability, verification, and legal judgment remain with the human legal professional.