A Court Appeal Collapsed Because the Cited Case Was Invented by a Chatbot
What happened
An appeal is a high-stakes document. It goes before federal judges, it cites precedent, and it is supposed to be grounded in actual law. When New York attorney Jae Lee filed an appeal in January 2024 to revive her client's lawsuit, she included a case that appeared to do exactly what she needed: it described a Queens doctor botching an abortion and established relevant precedent. The problem was that the case did not exist. It had been produced by ChatGPT.
Lee had used the chatbot to conduct legal research, and the system returned case citations that looked authentic. The fabricated case was woven into her appeal for the 2nd US Circuit Court of Appeals without being verified against any actual court database. When the opposing side or the court's clerks went to look it up, it was not there. The appeal was dismissed on those grounds, and Lee's conduct was referred to the court's grievance panel.
The grievance panel found that Lee's conduct fell "well below the basic obligations of counsel." She now faces possible sanctions. The case at the center of the original litigation, Park v Kim, became secondary to the question of professional misconduct. An attorney who set out to revive a client's lawsuit ended up jeopardizing her own standing with the bar because she trusted AI-generated output as if it carried the same reliability as a verified case database.
Lee's case was one of several that surfaced around the same period. Other attorneys had done the same thing, citing nonexistent cases in federal and state court filings, and the pattern drew attention from bar associations and legal commentators watching generative AI expand into professional practice. The incidents collectively signaled that the legal profession's adoption of these tools had outpaced its protocols for using them.
The underlying problem is not what a language model is capable of fabricating. It is what a professional is required to verify before signing their name to a court document. Lawyers have always been responsible for the accuracy of their citations, and courts have always had the power to sanction those who file false ones. What changed is that a tool now exists that produces fabricated citations formatted to look indistinguishable from real ones. Nothing in Lee's workflow caught the error before the filing went out. A provable record of what a system produced, combined with a required verification step before it enters a legal document, would be the minimal infrastructure for closing that gap. Right now, no such requirement exists.
Reported impact
- Affected parties
- Not publicly disclosed
- Harm type
- Not publicly disclosed
- Scale
- Not publicly disclosed
- Financial impact
- Not publicly disclosed
- Regulatory action
- Not publicly disclosed
Classification
Relevant governance controls
Governance control mapping is not available for this record.
- No controls mapped
Not publicly disclosed
Control mapping is analytical. It does not state that any control would have prevented the incident.
Sources and evidence
This record was researched and written by the Index. The event is also catalogued in the following database, which is listed for cross-reference.