Korean judges encounter AI-generated fake legal precedents: Survey

Seoul: The majority of South Korean judges - eight out of ten - have encountered court cases where false or non-existent legal precedents generated by artificial intelligence (AI) were cited, according to a recent survey conducted by the Supreme Court and published on Monday, reported Yonhap News Agency. The survey, which polled 360 judges from October 6 to 16, aimed to examine the growing concern of AI hallucination - a term describing instances where AI models present false information as if it were true.

Of the judges surveyed, approximately 80% reported experiencing cases where nonexistent or false legal precedents were cited. Specifically, 196 judges, or 80.7%, mentioned encountering such citations during civil, family, and administrative cases, while 159 judges, or 65.4%, encountered courts citing false precedents.

Additionally, the survey revealed that 41 judges, or 16.9%, faced situations where false or non-existent laws and regulations were cited, and 10 judges, or 4.1%, reported seeing citations of false or non-existent theses and publications. In total, 164 judges admitted to having cited false precedents themselves, while 136, or 57.6%, stated that lawyers representing parties in court had cited false precedents or other misleading information.

The overwhelming majority, nearly 92%, of the respondents expressed the urgent need for institutional measures to deal with the issue of AI-generated fake legal precedents. Proposed measures include establishing legal grounds for imposing fines, setting verification requirements, and creating disciplinary actions against lawyers who repeatedly cite false information.

A representative from the National Court Administration highlighted that the citation of AI-generated false or nonexistent legal precedents could potentially cause delays in court proceedings, emphasizing the urgent requirement for effective countermeasures based on the survey's findings.