80% of South Korean judges encounter AI-generated fake legal precedents: Supreme Court survey
Synopsis
South Korea's Supreme Court has put a number on a problem most judiciaries are still reluctant to admit: eight in 10 judges have faced AI-generated fake legal citations in court. With 164 judges acknowledging they themselves cited false precedents, the crisis is inside the bench as much as before it — and nearly 92% are now demanding institutional action.
Key Takeaways
80 per cent of surveyed South Korean judges have encountered AI-generated false or non-existent legal precedents in court cases.
The survey covered 360 judges and was conducted between 6 and 16 October , with results published on 2 November 2026 .
196 judges ( 80.7% ) reported non-existent precedent citations; 159 judges ( 65.4% ) encountered outright false precedent citations.
164 judges admitted to personally citing false precedents; 136 judges ( 57.6% ) said lawyers had done the same.
Nearly 92 per cent of respondents called for institutional measures including fines, verification requirements, and disciplinary action.
The National Court Administration warned that AI-generated false citations risk causing significant delays in court proceedings .
South Korea's Supreme Court has revealed that eight in 10 judges have encountered court cases in which false or non-existent legal precedents generated by artificial intelligence (AI) were cited, according to a survey published on Monday, 2 November 2026. The findings, drawn from a poll of 360 judges, have intensified calls for institutional safeguards against what legal experts are calling a growing crisis of 'AI hallucination' in courtrooms.
Key Findings from the Survey
The survey, conducted between 6 and 16 October, focused on judges presiding over civil, family, and administrative cases. Of those polled, 196 judges, or 80.7 per cent, reported encountering citations of non-existent legal precedents in court proceedings, while 159 judges, or 65.4 per cent, said they had dealt with cases involving citations of outright false precedents.
The problem extended beyond case law. 41 judges, representing 16.9 per cent of respondents, encountered citations of false or non-existent laws and regulations, while 10 judges, or 4.1 per cent, reported seeing fabricated theses and academic publications cited in proceedings.
Judges and Lawyers Both Implicated
Notably, the survey found the issue is not confined to one side of the courtroom. A total of 164 judges acknowledged having themselves cited false precedents, while 136 judges, or 57.6 per cent, said lawyers representing parties in court had cited false precedents or other AI-generated inaccurate information. This dual accountability — spanning both the bench and the bar — makes the challenge structurally more complex to address.
What Judges Are Demanding
Nearly 92 per cent of respondents said institutional measures are urgently needed to curb AI-generated citation errors. Proposed countermeasures cited in the survey include establishing legal grounds for imposing financial fines, introducing mandatory verification requirements before submissions, and crafting disciplinary mechanisms targeting lawyers who repeatedly cite false information.
An official at the National Court Administration warned that the citation of AI-generated false or non-existent precedents risks causing significant delays in court proceedings, underscoring the urgency of crafting effective policy responses grounded in the survey's data.
The Broader 'AI Hallucination' Problem
The phenomenon of AI hallucination — where large language models confidently present fabricated information as established fact — has become a global concern in professional settings, with legal systems among the most acutely exposed. This comes amid rising adoption of AI-assisted legal research tools in courtrooms across Asia, North America, and Europe. South Korea's findings mirror incidents reported in the United States, where lawyers have faced sanctions for submitting AI-generated citations of non-existent cases. The Supreme Court's proactive survey signals that Seoul is among the first judiciaries in Asia to formally quantify the scale of the problem.
With nearly nine in ten judges backing institutional intervention, South Korea's legal system now faces a critical window to establish precedent-setting frameworks that other judiciaries may look to as a model.
Point of View
But the proposed remedies — fines and disciplinary notices — address symptoms, not the root cause: legal research tools that generate plausible-sounding falsehoods with no built-in verification layer. Until AI providers and bar associations co-design citation-verification standards, courtrooms across the world remain exposed to the same risk, regardless of how many sanctions are threatened after the fact.
NationPress
2 Nov 2026
Frequently Asked Questions
What did South Korea's Supreme Court survey on AI fake legal precedents find?
The survey found that about 80 per cent of South Korean judges have encountered court cases where false or non-existent legal precedents generated by AI were cited. Conducted on 360 judges between 6 and 16 October, the findings were published on 2 November 2026.
What is 'AI hallucination' and why does it matter in courts?
AI hallucination refers to a situation where AI models present fabricated information as if it were factual. In a legal context, this can mean an AI-powered research tool generating citations to cases, laws, or academic publications that do not actually exist, potentially misleading judges and delaying proceedings.
Were judges themselves responsible for citing AI-generated false precedents?
Yes. A total of 164 judges in the survey acknowledged having personally cited false precedents, while 136 judges — or 57.6 per cent — also reported that lawyers representing parties had cited false or non-existent information in their courts.
What measures are South Korean judges calling for?
Nearly 92 per cent of respondents said institutional measures are needed. These include establishing legal grounds for imposing financial fines, introducing mandatory verification requirements before submissions, and setting up disciplinary mechanisms against lawyers who repeatedly cite AI-generated false information.
Is AI-generated fake legal citation a problem in other countries too?
Yes. Similar incidents have been reported in the United States, where lawyers faced court sanctions for submitting AI-generated citations to non-existent cases. South Korea's formal survey is among the first in Asia to quantify the scale of the problem, potentially setting a benchmark for other judiciaries.
