In brief
A recent Supreme Court survey shows that 80.7% of surveyed judges reported cases where AI‑generated non‑existent precedents were cited, and 65.4% noted false citations. 92% of respondents urged the creation of fines, verification rules and disciplinary actions to combat the growing issue of AI hallucination in Korean courts.
In a study released Monday, the Supreme Court of South Korea announced that about eight out of ten judges had encountered court cases citing fake or non‑existent legal precedents generated by artificial intelligence.
Survey Details
The survey, conducted from October 6‑16 and covering 360 judges who handle civil, family and administrative matters, found that 196 judges—80.7% of the total—had seen citations of non‑existent precedents in court filings. 159 judges—65.4%—reported seeing cases where the AI‑generated citations were outright false.
Less common but still significant were instances where AI produced inaccurate laws or regulations. 41 judges—16.9%—encountered citations of false or non‑existent statutes, while 10 judges—4.1%—reported the presence of fabricated theses and scholarly publications in legal arguments.
The survey also recorded the source of the misinformation. 164 judges—57.6%—said they had personally cited false precedents, while 136 judges—over one‑third of respondents—stated that lawyers on either side of the case had introduced erroneous information into their arguments.
Nearly 92% of the judges surveyed demanded institutional action to curb the practice. Proposed measures include setting legal grounds for imposing fines, establishing verification requirements for all AI‑generated content, and devising disciplinary penalties for lawyers who repeatedly rely on false information.
An official at the National Court Administration warned that the use of AI‑fabricated precedents could delay court proceedings and urged the judiciary to adopt effective countermeasures based on the survey results.
