Mata v. Avianca, Inc., 678 F.Supp.3d 443

Background

Plaintiff Roberto Mata sued Avianca for injuries under the Montreal Convention; case removed to federal court.
Two Levidow Firm lawyers were involved: Steven A. Schwartz (researched and drafted) and Peter LoDuca (filed and signed papers; not admitted in the district).
Schwartz used ChatGPT for legal research; ChatGPT generated fabricated judicial opinions and quoted fake language and citations.
LoDuca filed a March 1 affirmation (signed under penalty of perjury) citing those authorities and later filed an April 25 affidavit that annexed excerpts of the fabricated opinions; he did not meaningfully verify the authorities.
Avianca’s reply (Mar. 15) and the Court’s Orders (Apr. 11–12) flagged that multiple cited cases could not be located; respondents delayed full disclosure and only acknowledged ChatGPT’s fabrication after an Order to Show Cause.
The Court found subjective bad faith (conscious avoidance and misleading statements), imposed Rule 11 and inherent-power sanctions against LoDuca and Schwartz, and held the Levidow Firm jointly liable.

Issues

Issue	Mata (Plaintiff) — position	Avianca (Defendant) — position	Held
Whether citing and submitting non-existent judicial opinions violated Rule 11 and warranted sanctions	Argued opposition to dismissal relying on cited authorities (affirmation presented as true)	Argued many authorities cited in plaintiff’s papers do not exist and urged sanctions	Court: citations were fabricated by ChatGPT; Rule 11 and inherent-power sanctions appropriate for bad-faith submission and later advocacy
Whether LoDuca personally acted improperly in signing filings and failing to verify authorities	Claimed he relied on colleague Schwartz and believed submissions were accurate	Avianca emphasized LoDuca did not read opponent’s reply, did not verify cases, lied about being on vacation to get an extension	Court: LoDuca acted in subjective bad faith (no meaningful inquiry, false vacation claim, swore to affidavit without basis)
Whether Schwartz’s use of ChatGPT excuses responsibility or demonstrates misconduct	Claimed ChatGPT was a research “supplement” and he was misled by the tool	Avianca argued Schwartz relied on ChatGPT to create fake opinions and failed to verify results	Court: Schwartz acted in subjective bad faith (conscious avoidance, misleading statements about reliance); fabricated cases admitted to be ChatGPT-generated
Whether the law firm is jointly liable and the scope of sanctions (monetary and non-monetary)	Firm acknowledged errors, implemented training, and offered remediation steps	Avianca sought corrective measures and sanctions; did not press for fees	Court: Levidow Firm jointly and severally liable (no exceptional circumstances); ordered letters to affected judges and plaintiff, and a $5,000 penalty to the Court registry

Key Cases Cited

Cooter & Gell v. Hartmax Corp., 496 U.S. 384 (1990) (Rule 11 purpose: encourage reasonable inquiry and deter abusive filings)
Chambers v. NASCO, Inc., 501 U.S. 32 (1991) (courts’ inherent power to sanction for abuse of the judicial process)
Muhammad v. Walmart Stores East, L.P., 732 F.3d 104 (2d Cir. 2013) (sua sponte Rule 11 sanctions require subjective bad faith)
Salovaara v. Eckert, 222 F.3d 19 (2d Cir. 2000) (abuse of adversary system supports sanctions for frivolous arguments)
Storey v. Cello Holdings, L.L.C., 347 F.3d 370 (2d Cir. 2003) (distinguishing mere error from sanctionable conduct under Rule 11)
Fishoff v. Coty Inc., 634 F.3d 647 (2d Cir. 2011) (long‑shot theories not necessarily sanctionable; standard for frivolous legal contentions)
In re 60 E. 80th St. Equities, Inc., 218 F.3d 109 (2d Cir. 2000) (bad faith may be inferred when claims are completely without merit)
United States v. Reich, 479 F.3d 179 (2d Cir. 2007) (forgery of judicial signatures implicates integrity of federal judicial proceedings)
United States v. Svoboda, 347 F.3d 471 (2d Cir. 2003) (conscious avoidance can satisfy knowledge standard for sanctions)
United States v. Finkelstein, 229 F.3d 90 (2d Cir. 2000) (defining conscious avoidance test)