NEW YORK — Two apologetic attorneys who responded to an angry Manhattan federal court judge blamed ChatGPT on Thursday for tricking them into including bogus legal research in a court filing.
Attorneys Stephen A. Schwartz and Peter LoDuca are facing potential punishment for filing a lawsuit against an airline that included references to previous court cases that Schwartz thought were real, but were actually invented by the AI-powered chatbot.
Schwartz said he used the groundbreaking software to search for legal precedents supporting a client’s case against Colombian airline Avianca over a 2019 in-flight injury.
The chatbot, which has dazzled the world by producing essay-like answers to claims from users, has suggested several cases involving airline accidents that Schwartz couldn’t find through the usual methods used at his law firm.
The problem was that many of these cases were not real or involved non-existent airlines.
Schwartz told US District Judge B. Kevin Castle stated that he “was operating under the misconception… that this site was getting these issues from some sources that I didn’t have access to.”
He said he “failed miserably” in conducting follow-up research to ensure the citations were correct.
“I didn’t understand that ChatGPT could make issues up,” Schwartz said.
Microsoft has invested around $1 billion in OpenAI, the company behind ChatGPT.
Its success, which has shown how artificial intelligence can change the way humans work and learn, has generated concerns for some. Hundreds of industry leaders signed a letter in May warning that “mitigating the extinction risk from AI should be a global priority along with other societal risks such as pandemics and nuclear war.”
Judge Castell appeared baffled and upset by the unusual event and disappointed that attorneys did not act quickly to correct the false legal citations when they were first alerted to the problem by Avianca’s attorneys and the court. Avianca cited spurious case law in the March filing.
The judge confronted Schwartz with a single legal case that the computer program had invented. It was initially described as a wrongful death case brought by a woman against an airline only to turn into a legal action about a man who missed a flight to New York and had to incur additional expenses.
“Can we agree that this is legal nonsense?” Castell asked.
Schwartz said he mistakenly believed the confusing presentation was caused by excerpts taken from different parts of the case.
When Castel finished questioning him, Schwartz asked if he had anything else to say.
“I would like to sincerely apologize,” Schwartz said.
He added that he had suffered personally and professionally as a result of the blunder and felt “embarrassed, humiliated and deeply remorseful”.
He said that he and the company he worked for — Livido, Livido and Opperman — had put safeguards in place to ensure nothing similar would happen again.
Luduca, the other attorney who worked on the case, said he trusted Schwartz and had not adequately reviewed what he had collected.
After the judge read parts of one of the cases cited aloud to show how easy it was to discern that it was “gibberish,” Loduca said, “It never occurred to me that this was a bogus case.”
He said the result “hurts me to no end”.
Ronald Minkoff, an attorney for the law firm, told the judge the request “caused negligence, not bad faith” and should not result in penalties.
Lawyers have historically had a hard time with technology, especially new technology, he said, “and it hasn’t gotten any easier.”
Mr. Schwartz, who barely does federal research, chose to use this new technology. Minkoff said he thought he was dealing with a standard search engine, and what he was doing was playing with live fire.
Daniel Shen, associate professor and assistant director of research in the Center for Legal and Court Technology at the College of William & Mary Law, said he introduced Avianca’s case during a conference last week that drew dozens of participants in person and online from state and federal. Courts in the United States, including the Manhattan Federal Court.
He said that the subject caused shock and confusion in the conference.
“We’re talking about the Southern District of New York, the federal district that deals with the big cases, 9/11 for all the big financial crimes,” Sheen said. “This was the first documented case of possible professional misconduct by a lawyer using generative AI.”
He said the case demonstrated how lawyers may not understand how ChatGPT works because it tends to hallucinate, talking about fictional things in a way that seems realistic but is not.
“It highlights the dangers of using promising AI technologies without knowing the risks,” Shen said.
The judge said that he will rule on the penalties at a later time.
“Reader. Infuriatingly humble coffee enthusiast. Future teen idol. Tv nerd. Explorer. Organizer. Twitter aficionado. Evil music fanatic.”