In brief
A recent survey by the Korean Supreme Court revealed that 80 % of judges have encountered cases citing non‑existent or false AI‑generated legal precedents, with many citing legal or procedural safeguards needed. The study underscored the risk of AI hallucination in courts and the necessity for institutional measures. The survey, covering 360 judges, highlighted widespread use of fabricated precedents and raised calls for formal penalties and verification protocols. However, questions remain about the underlying causes of this phenomenon and the future effectiveness of proposed reforms.
In a comprehensive nationwide survey released by South Korea’s Supreme Court, 80 % of judges reported that they had faced cases where artificial‑intelligence‑generated legal precedents—actual or false—were cited in court proceedings, uncovering a troubling trend in the use of AI‑hallucinatory data within the judiciary.
Survey Findings
The survey, which interviewed 360 judges who presided over civil, family, and administrative matters between October 6 and 16, showed that 196 judges, or 80.7 percent, had seen citations of nonexistent precedents, while 159 judges, 65.4 percent, reported encounters with false precedents. The study also recorded that 41 judges, 16.9 percent, had encountered false or non‑existent laws and regulations, and 10 judges, 4.1 percent, noted misuse of fabricated theses and publications. In total, 164 judges admitted to citing false precedents themselves, and 136 judges, 57.6 percent, documented that lawyers representing parties had cited such misinformation.
Calls for Institutional Safeguards
Nearly 92 percent of respondents said that institutional measures must be introduced to counter the citations of AI‑generated fake precedents. Suggested interventions include legal provisions for imposing fines, mandatory verification protocols for precedent claims, and disciplinary actions against attorneys who repeatedly present fabricated information. An official at the National Court Administration warned that the use of false AI‑generated data can delay proceedings, stressing the need for countermeasures informed by the survey’s findings.
The researchers highlighted the phenomenon of “AI hallucination”—a situation where models present false information as if it were true—emphasizing that it is increasingly threatening to legal integrity. By exposing the prevalence of fabricated precedents, the survey also calls into question how judicial systems anticipate and mitigate emerging AI‑related risks. The spread of advanced language models that can produce plausible yet non‑existent legal citations underscores a necessity for both safeguards and ongoing vigilance across the legal community.
Though the Supreme Court has begun to outline a roadmap for reforms, exact timelines and legislative routes remain unclear, and further analysis is required to determine the best balance between technological innovation and courtroom reliability.
In light of these findings, the official response focuses on the enforcement of existing sanctions and the creation of new verification practices, but precise details for the implementation of fines, legal deadlines, and verification standards have yet to be finalized.
Ultimately, the uncertainty lies in how quickly the judiciary can adapt to the rapid advancement of AI technologies and whether the proposed measures will be sufficient to prevent the misuse of fabricated legal precedents.
