Mata v. Avianca, Inc.

Background

Roberto Mata sued Avianca, Inc. after he alleged that a metal serving cart struck his knee during a flight from El Salvador to John F. Kennedy Airport. Avianca moved to dismiss, arguing that Mata’s claims were barred by the Montreal Convention’s two-year limitations period.

In opposition to that motion, the Levidow Firm submitted an affirmation citing and quoting purported judicial decisions. The court later determined that several cited decisions did not exist and that ChatGPT had generated them. The submission also relied on existing authorities that generally did not support the propositions for which they were cited, with limited exceptions identified by the court.

Steven A. Schwartz researched and drafted the opposition. Peter LoDuca, who was the attorney of record in federal court, signed and filed it without reviewing any of the cited authorities. After Avianca questioned the citations, neither lawyer withdrew the submission or explained how the cited decisions could not be found. The court then ordered LoDuca to provide copies of the cited decisions. The response included excerpts of purported opinions rather than complete opinions and stated that another cited decision could not be located.

Findings and legal conclusions

The court found that Schwartz used ChatGPT as the source of his substantive research after finding that the firm’s research service provided insufficient access to federal cases. He prompted ChatGPT for supporting case law, and the chatbot generated descriptions and excerpts of nonexistent cases. Schwartz did not have the full text of any decision generated by ChatGPT when he prepared the opposition.

The court found subjective bad faith, meaning knowing or consciously avoided wrongdoing, by LoDuca and Schwartz. LoDuca failed to make any meaningful inquiry into the authorities before signing the opposition, did not read Avianca’s reply before forwarding it to Schwartz, and later signed an affidavit without a basis for confirming its truth. The court also found that LoDuca knowingly made a false statement when he sought an extension by saying he was away on vacation, when Schwartz was the person who was away.

The court found that Schwartz knew or had reason to suspect that at least some cited cases did not exist but did not disclose that information. It also found that his descriptions of ChatGPT as merely supplementing his research, and his explanations about when he questioned whether the cases were real, were misleading or contradictory. The court concluded that submitting and continuing to advocate fake opinions violated Federal Rule of Civil Procedure 11.

The court held that the Levidow Firm was jointly and severally responsible for LoDuca’s and Schwartz’s Rule 11 violations because the firm had not shown exceptional circumstances. The court declined to impose a separate sanction under 28 U.S.C. § 1927 because the conduct did not involve unreasonable and vexatious multiplication of proceedings. The court also concluded that the respondents had not violated 18 U.S.C. § 505 because the fake opinions did not contain a forged judicial signature or court seal.

Sanctions and order

Judge Castel ordered the respondents to send, within 14 days, individually addressed letters to Mata and to each judge falsely identified as the author of the fake opinions. The letters had to identify and include this opinion, the June 8, 2023 hearing transcript, and the April 25 affirmation with its exhibits. The respondents also had to file copies of those letters with the court.

The court jointly and severally imposed a $5,000 penalty on LoDuca, Schwartz, and the Levidow Firm, payable into the court’s registry within 14 days. The court did not order payment of Avianca’s attorneys’ fees or expenses and did not require an apology or additional educational sanctions.