Law Firm Sanctioned $5,000 After ChatGPT Fabricates Six Court Cases in Legal Brief

Attorneys at Levidow, Levidow & Oberman needed case-law precedent to oppose Avianca Airlines' motion to dismiss a personal injury suit brought by client Roberto Mata. Attorney Steven Schwartz, who had practiced in litigation for approximately 30 years, had not previously used ChatGPT. The firm used Fastcase, whose account had limited access to federal cases, and did not maintain Westlaw or LexisNexis accounts. Schwartz testified that he "falsely assumed" ChatGPT "was like a super search engine" and used it for research.

Schwartz used ChatGPT to find and draft the citations supporting the brief. When he questioned ChatGPT about whether "Varghese" was a real case and whether the other cases it provided were fake, ChatGPT responded that it had supplied "real" authorities that could be found through Westlaw, LexisNexis and the Federal Reporter. Asked by the court whether he did any other research, Schwartz said no, other than initially going to Fastcase and failing there. Co-counsel Peter LoDuca signed and filed the brief, which Schwartz had researched and written, without reviewing any of the cited authorities.

Six decisions cited in the brief, "Varghese", "Miller", "Petersen", "Shaboon", "Martinez" and "Durden", were generated by ChatGPT and do not exist, with fake quotes and citations. Judge Kevin Castel's June 22, 2023 opinion found bad faith on the part of Schwartz and LoDuca, based on conscious avoidance and false and misleading statements to the court after the cases were questioned. The court imposed a $5,000 penalty, jointly and severally, on Schwartz, LoDuca and the Levidow firm, payable into the registry of the court. Separately, the court granted Avianca's motion to dismiss Mata's underlying suit as time-barred under the Montreal Convention.

• Generative AI can produce non-existent judicial opinions with fake quotes and citations; here ChatGPT also told the lawyer, when he asked, that the authorities were real.
• AI-generated research output must be independently verified before it is relied on or filed; the court said Rule 11 gives attorneys a gatekeeping role to ensure the accuracy of their filings.
• The court found bad faith based on conscious avoidance and false and misleading statements after the cases were questioned, and imposed a penalty on both lawyers and their firm.