Anger, in writing

The claim under review holds that a court confronted filings produced by generative AI that invented content — described in Korean shorthand as an "AI novel" — and that the judges responded by putting their reproach directly into the judgment text: the phrase "false judgment," an accusation of "deceiving the court," and a finding of "serious illegality." The episode has since been cited as an example of the risks posed by unverified output from AI systems, including LLMs, entering official court records.

What the review found

Reviewers examined three primary sources. The decisive one is training material prepared for judges at the Judicial Research and Training Institute, dated September 4, 2026, which is cited as the basis for the assessment. That the episode appears in material used for judicial training indicates it is being treated as a cautionary reference on AI fabrication within the judiciary itself.

The gaps

Because the corroborating document is non-public, details that would normally let readers verify the story — the court, the case, the parties, and the full wording of the ruling — cannot be checked against open records. The review found zero publicly accessible web sources documenting the training material, meaning the assessment rests on documents available to the reviewers rather than to the general reader, and the absence of a public trail leaves room for residual doubt.

On the strength of the non-public training record and the lack of any contrary evidence, the claim that judges provoked by an "AI novel" filing wrote "false judgment," "deceiving the court" and "serious illegality" into their ruling is Mostly True.