Abstract
The common law, originating in England and later adopted in countries such as the United States, Canada, Australia, India, Hong Kong, and Singapore, remains unified by the doctrine of stare decisis, the binding force of precedent. Accurate citation of case law is, therefore, an essential skill for judges and lawyers, particularly in case analysis and legal drafting. Despite efforts to ensure reliable training data, such systems may generate inaccuracies—misleadingly referred to as "hallucinations." Unlike human error, these are statistical variances rather than cognitive phenomena. In common law systems, however, reliance on false precedents undermines the authority of binding precedent, since it is no longer possible to determine whether a precedent is genuine and how it connects to the actual judicial decision that gave rise to the principle of stare decisis. The purpose of this paper is to examine the potential consequences of using Large Language Model (LLM) in legal practice, and to consider possible safeguards, with particular regard to the preservation of the binding authority of leading cases within common law jurisdictions.