Incident Tracker / United States (federal)
Mata v Avianca, Inc.
The origin case of legal AI hallucination sanctions: two New York attorneys were fined $5,000 after citing six ChatGPT-invented cases in a personal-injury filing and persisting after the fabrications were questioned, prompting the first widely reported judicial rebuke of generative AI misuse by lawyers.
- Court
- US District Court, Southern District of New York
- Date
- 2023-06-22
- Who used the AI
- Not recorded
- AI tool
- ChatGPT
- Conduct
- Attorneys filed an affirmation citing six non-existent judicial decisions generated by ChatGPT, then doubled down when challenged, submitting fabricated opinions.
- Outcome
- Sanctions
- Penalty
- US$5,000
- Regulatory outcome
- No separate regulatory disposition recorded
- Status
- Verified against listed sources
- Last checked
- 2026-09-16
What was fabricated or misused
- Varghese v. China Southern Airlines Co., Ltd., 925 F.3d 1339 (11th Cir. 2019) — generated by ChatGPT and did not exist
- Shaboon v. Egyptair, 2013 IL App (1st) 111279-U (Ill. App. Ct. 2013) — generated by ChatGPT and did not exist
- Peterson v. Iran Air, 905 F. Supp. 2d 121 (D.D.C. 2012) — generated by ChatGPT and did not exist
- Martinez v. Delta Airlines, Inc., 2019 WL 4639462 (Tex. App. Sept. 25, 2019) — generated by ChatGPT and did not exist
- Estate of Durden v. KLM Royal Dutch Airlines, 2017 WL 2418825 (Ga. Ct. App. June 5, 2017) — generated by ChatGPT and did not exist
- Miller v. United Airlines, Inc., 174 F.3d 366, 371-72 (2d Cir. 1999) — generated by ChatGPT and did not exist
What the court said
“Respondents abandoned their responsibilities when they submitted non-existent judicial opinions with fake quotes and citations created by the artificial intelligence tool ChatGPT, then continued to stand by the fake opinions after judicial orders called their existence into question.”
“Respondents have now acknowledged that the “Varghese”, “Miller”, “Petersen”, “Shaboon”, “Martinez” and “Durden” decisions were generated by ChatGPT and do not exist.”
“A fake opinion is not “existing law” and citation to a fake opinion does not provide a non-frivolous ground for extending, modifying, or reversing existing law, or for establishing new law.”
“A penalty of $5,000 is jointly and severally imposed on Respondents and shall be paid into the Registry of this Court within 14 days of this Opinion and Order.”
Timeline
- Avianca moved to dismiss, arguing the Montreal Convention time bar.
- LoDuca filed the opposition affirmation citing purported decisions.
- Avianca's reply said most cited cases could not be located.
- Court ordered LoDuca to attach copies of seven cited decisions.
- LoDuca filed excerpts purporting to be the required decisions.
- Court held the sanctions hearing.
- Court imposed a $5,000 joint penalty and nonmonetary directives.
Why this case matters
Mata is the tracker’s baseline for lawyer-filed AI hallucinations. The order matters because the sanction did not rest only on the first ChatGPT error; it rested on counsel's failure to verify, false statements about how the research was done, and continued reliance after Avianca and the court questioned the cases. Later records such as Mezu and Douglas repeat that distinction: using AI is not forbidden, but signing or defending fake authority after a reasonable check would have exposed it triggers court action.
Practice note
If AI suggests a case, treat it as unverified until the full decision has been opened in an official reporter, court database, Westlaw, Lexis, or another reliable source. If an opponent or judge says a citation cannot be found, stop relying on it, disclose the research path, and correct the filing before supplying excerpts or summaries.
Primary sources for this incident
Related regulation records for this incident
- ABA Formal Opinion 512: Generative Artificial Intelligence ToolsABA · US-FED
- Report and Recommendations of the New York State Bar Association Task Force on Artificial IntelligenceNYSBA · US-STATE
- Standing Order Re: Artificial Intelligence in Cases Assigned to Judge BaylsonE.D. Pa. (Baylson) · US-FED
- Mandatory Certification Regarding Generative Artificial Intelligence (Judge Brantley Starr)N.D. Tex. (Starr) · US-FED
Similar incidents on the record
- Johnson v. Dunn, No. 2:21-cv-01701-AMM (N.D. Ala.), 792 F. Supp. 3d 12412025-07-23 · US District Court, Northern District of Alabama (Southern Division) · SanctionsIn the Northern District of Alabama, Judge Anna M. Manasco publicly reprimanded Butler Snow attorneys Matthew B. Reeves, William J. Cranford and William R. Lunsford for filing two motions containing citations hallucinat…
- State v. Coleman, 2026-Ohio-965 (Ohio Ct. App., 11th Dist.)2026-03-20 · Ohio Court of Appeals, Eleventh Appellate District · SanctionsAn Ohio appellate court fined a criminal-appeal attorney $2,000, removed him as counsel, struck his reopening application, and referred him to the Ohio Office of Disciplinary Counsel after his staff's ChatGPT use produc…
- Robert Dastou v. Kathryn Holmes, No. 2381CV02212 (Mass. Super. Ct.)2025-07-16 · Massachusetts Superior Court · SanctionsA Massachusetts Superior Court sanctioned a defense attorney who used ChatGPT to 'format' filings, producing motions in limine and jury instructions riddled with hallucinated cases, a fake quotation, and a fictitious le…
- Garner v. Kadince, Inc., 2025 UT App 802025-05-22 · Utah Court of Appeals · SanctionsIn the Utah Court of Appeals, petitioner's counsel filed a brief containing ChatGPT-fabricated citations, including a non-existent 'Royer v. Nelson', violating appellate rule 40. The court sanctioned attorney Richard Be…
Cite this record
SafeLegalAI Global Legal AI Incident Tracker, “Mata v Avianca, Inc.”, safelegalai.com/tracker/mata-v-avianca (accessed 2026-09-16). Data: CC BY 4.0.