PreviewYou're previewing CLO AI Chambers. Sign in to use it.Sign in
← All lessons
L2 · Confidentiality, privilege, accuracy and professional duties8 min read

Verification: never file a fake citation

The most visible legal AI failure is the fake citation: a brief or submission that cites cases that don't exist, or quotes words a court never wrote. It has happened in hundreds of cases worldwide, and courts treat it as a serious breach of the duty of candour, whatever tool caused it.

The landmark cases

  • Mata v. Avianca (S.D.N.Y. 2023): lawyers filed an opposition citing non-existent decisions generated by ChatGPT, then defended them when questioned. Judge P. Kevin Castel sanctioned the two lawyers and their firm, jointly fining them $5,000, and found bad faith in how they responded.
  • Ayinde v London Borough of Haringey (High Court of England and Wales, 2025): a Divisional Court led by the President of the King's Bench Division dealt with two cases of fictitious citations, made wasted-costs orders and referred lawyers to their regulators, stressing that lawyers who use AI must check its output against authoritative sources.
  • Johnson v. Dunn (N.D. Ala. 2025): lawyers from a large firm were disqualified from the case after filings with AI-fabricated citations, despite the firm's internal AI warnings.

A verification workflow that works

  1. Pull every authority (case, statute, regulation, quotation) into a list.
  2. Check each one in a primary source: it exists, the citation is right, the quote is verbatim, the proposition is supported and it is still good law.
  3. Record who checked and when.
  4. Second-person check for filings to courts and regulators.
  5. Check local rules: some judges require certification of AI use or verification.

Scale the review to the risk

Not every AI output needs a cite-check. An internal summary for a lawyer's own use needs a spot-check; advice to the board needs full review; a court filing needs full review plus a recorded cite-check.

What this means for the General Counsel

What this means for a General Counsel

Your name may not be on the brief, but your company's is. Make verification a requirement of your outside counsel guidelines and a standard step for your own team's filings, regulator submissions and board papers.

Ask your firms how they verify AI-assisted work. Courts have sanctioned large firms as well as small ones; a policy alone is not enough, as Johnson v. Dunn showed. What matters is a recorded check by a named person.

Diagram

Verification: never file a fake citation

Real example

In Mata v. Avianca, Inc. (S.D.N.Y.), lawyers for the plaintiff filed an affirmation citing judicial decisions that did not exist, generated by ChatGPT. On 22 June 2023 Judge P. Kevin Castel imposed sanctions on two lawyers and their firm, including a $5,000 penalty imposed jointly and severally, finding they had acted in subjective bad faith by continuing to stand by the fake opinions after they were questioned. Source: Wikipedia — Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023) ↗ · Reed Smith — ChatGPT artificial intelligence tool generates fake opinions (Mata v. Avianca, S.D.N.Y. June 22, 2023) ↗

Use it

Quick check · 3 questions

  1. 1.What does a proper cite-check confirm for each AI-supplied authority?

  2. 2.Why did the court in Mata v. Avianca find bad faith?

  3. 3.Which output needs the strictest verification?

0 of 3 answered

Sources

Last reviewed 2026-10-01.