🔍 Read the full analysis: Fake Precedents And AI In Korea: A Survey-Based Look At Judges’ Views on ThorstenMeyerAI.com
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
South Korea’s National Court Administration said a survey of 360 judges found frequent encounters with false or distorted legal citations. The reported figures describe judges’ experiences, not the share of court filings containing fake material; the administration is considering procedural-law changes, including fines.
South Korea’s National Court Administration published results on November 2 from a survey of 360 judges, reporting that many judges handling civil, family and administrative cases had encountered fabricated or distorted legal citations in court filings. The findings include nonexistent precedents and real rulings described inaccurately, and have prompted calls for rules on citation checks and AI-use disclosure.
The survey was conducted from October 6 to 16 among judges at courts of all levels. The court administration said about 80% of responding judges in civil, family and administrative cases had encountered filings citing false or nonexistent precedents. That figure applies to a particular group of respondents, not to all judges or to the proportion of filings containing fabricated citations.
Among judges who reported encountering false citation material, 196 (80.7%) identified precedents that did not exist, while 159 (65.4%) identified real precedents cited with false content. Respondents could select more than one type. A separate report by the Korean daily Kyunghyang Shinmun put the number of judges across the full survey who had encountered fake citations at 243 of 360, or 67.5%; that overall figure is not printed in the court administration’s release.
Judges also identified who had submitted the material. 164 respondents (69.5%) selected litigants acting on their own, and 136 (57.6%) selected lawyers representing or defending parties. Those responses are not mutually exclusive. The Korean release says false citations appeared in documents submitted by both parties and lawyers. An English-language version carried by The Korea Herald, based on a Yonhap report, described the 164 judges as having cited false precedents themselves; the Korean original instead identifies litigants as the source.
Pressure for Citation Rules
The results point to a practical problem for court proceedings: judges and opposing parties may have to spend time checking whether cited authorities exist and whether a ruling actually supports the argument made. The court administration said false citations can create unnecessary verification work and hinder the speedy progress of trials. The survey records judges’ experiences; it does not measure delays or establish how often such filings occur across the justice system.
91.7% of respondents supported an institutional response, according to the release. The most-selected options were a legal basis for fines, chosen by 61.4%; a duty to verify citations, at 61.1%; and a duty to disclose AI use, at 60.6%. These are preferences expressed in a survey, not rules already in force. Any penalties or disclosure requirements would require further action.
Popular right now
Ninja Luxe Café Mini Plus Espresso and Drip Coffee Machine with AutoFroth, Stainless Steel (ES351)
Samsung 75-Inch M80H Series Mini LED 4K Smart TV (2026)
KitchenAid Artisan 5-Quart Tilt-Head Stand Mixer, Almond Cream
As an affiliate, we earn on qualifying purchases.
What the Survey Measured
The National Court Administration, an administrative body of South Korea’s Supreme Court, said the survey was prompted by concerns that AI systems can produce false or nonexistent legal authorities. Its stated concern was that such citations burden opposing parties with verification and can slow proceedings. The published release describes examples including invented judgments formatted like Supreme Court decisions, real case numbers attached to reasoning that could not be found, and genuine rulings quoted in ways that changed their meaning.
The administration also described filings that retained text from AI conversations or litigation-strategy notes. These examples show the range of material judges said they had encountered, but the release does not provide case names or enough detail to independently assess each incident. The survey was an account of judges’ observations, not an audit of court records or a study of the accuracy of particular AI systems.
The administration said it plans to propose changes to South Korea’s Civil and Criminal Procedure Acts to provide for fines. The source material also points to sanctions or referrals involving lawyers in New York and London, but the Korean survey itself concerns local judges’ reported experiences. The proposed amendments have not been enacted.
Limits of the Findings
The administration’s release does not state how many judges were invited, provide the wording of the survey questions, or report a response rate. It also does not count filings containing false citations. The percentages describe responses to particular questions, and several questions allowed multiple answers, so categories may overlap.
The release’s roughly 80% figure concerns responding judges assigned to civil, family and administrative cases. The 67.5% figure for all respondents comes from Kyunghyang Shinmun’s reported count of 243 judges and is not directly printed in the release. Further details about the cases, how the cited material was verified, and whether AI was used in each instance are not supplied.
It is also unclear when the administration will submit its proposed amendments, what exact conduct would trigger a fine, or whether any disclosure duty would apply to every use of AI. The survey reports judges’ views; it does not establish that the proposed measures will become law.
Proposed Court Procedure Changes
The next stated step is for the National Court Administration to propose amendments to the Civil and Criminal Procedure Acts that would allow fines for false citations. The release does not give a timetable or draft language. Until lawmakers act, the survey’s preferred measures—including citation-verification and AI-disclosure duties—remain proposals rather than established requirements.
Any legislative proposal will need to clarify how courts would distinguish deliberate misrepresentation from an accidental error, how citations should be checked, and what process would apply before a penalty. The administration has not supplied those details in the published survey results. Further updates are expected if the proposed changes are formally submitted or debated.
Key Questions
How many judges took part in the survey?
360 judges from courts at all levels responded. The survey ran from October 6 to 16, 2026, and the National Court Administration published the results on November 2.
Does the survey show that 80% of court filings contain fake citations?
No. The roughly 80% figure concerns responding judges handling civil, family and administrative cases who said they had encountered false or nonexistent precedents. The survey does not measure the share of filings containing fake material.
What kinds of false citations did judges report?
The most commonly selected categories were precedents that did not exist and real precedents cited with false content. The court administration also described false or nonexistent statutes and legal literature, as well as AI-generated text left in filings.
Did judges say lawyers submitted false citations?
Yes. 136 respondents (57.6%) selected lawyers acting for a party, while 164 selected litigants acting on their own. The categories could overlap. The Korean release does not say that 164 judges admitted citing false precedents themselves.
Will South Korea impose fines for false citations?
Not on the basis of this survey alone. The National Court Administration said it plans to propose legal changes that would allow fines, but the results do not provide a timetable or indicate that amendments have been enacted.
Source: ThorstenMeyerAI.com
