ChatGPT’s meteoric rise is shaking up many industries — including law, as one lawyer recently discovered.
Roberto Matta sued Avianca Airlines over injuries to his stroller during a 2019 flight, claiming the employee was negligent. Steven Schwartz, an attorney with Levidow, Leviw & Oberman and licensed in New York for more than three decades, represented Mattan.
But in at least six of the cases brought by Schwartz, Judge Kevin Casteel of the Southern District of New York ordered that “false citations and false citations appear to be false jury verdicts.”
Source of false cases? Discussion GPT
“The court is presented with unprecedented precedent,” Casteel wrote in the May 4 order.
Among the cases cited: Varghese v. China Southern Airlines, Martinez v. Delta Air Lines, Shabon v. EgyptAir, Peterson v. Iran Air, Miller v. United Airlines, and Durden v. KLM Royal Dutch Airlines Estate, all did. “It didn’t seem like it would work for the judge or the defense,” the record said.
Schwartz, in an affidavit, said he had never used ChatGPT as a legal research resource prior to this lawsuit, so he “was not aware that the content might be false.” He took responsibility for not verifying the chatbot’s sources.
Schwartz now faces a sentencing hearing on June 8.
In his affidavit this week, he said, “I am deeply concerned about the use of artificial intelligence to supplement the legal studies that have been conducted in this area and will never do so without fully verifying its accuracy in the future.”
In late April, Avianca lawyers from Condon & Forsyth wrote to Castells, questioning the validity of the cases.
In an affidavit filed Thursday, fellow attorney Peter Loducka said he had “no reason to question the integrity” of Schwartz’s research and had no role in the study.
In an affidavit filed on April 25, Schwartz was ordered to show cause not to be convicted of “false and fraudulent notarization.”
Schwartz’s affidavit Wednesday contained screenshots of the attorney’s claim to the ChatGPT case.
“It’s a matter of fact,” Schwartz asked the chatbot.
“Yeah,” ChatGPT doubled down, “it’s a real deal.”
Schwartz then asked for the source. The chatbot reiterated that the false case was true.
“I apologize in advance for any confusion,” replied ChatGPT. “After double checking, I found that the case Varghese v. China Southern Airlines Co., 925 F.3d 1339 (11th Cir. 2019), actually exists and can be found in legal research databases such as Westlaw and Lexis Nexis. I apologize for any inconvenience or confusion caused by my previous responses.
When Schwartz asked the chatbot if the other cases were false, ChatGPT replied that the other cases were “true” and could be found in “a popular legal database.”
CNN has reached out to Schwartz and LoDuca for comment.