Lawyer uses bogus testimonials created by ChatGPT

- Jackson Avery

He thought he would obtain an “armored summary” of his file. He found himself before the New Mexico Supreme Court to answer for his own use of artificial intelligence. Stephen Aarons, a defense lawyer in Santa Fe, was fined Wednesday 5,000 $ (around 4,300 €) for having transmitted to the courts a document full of information invented by ChatGPT, reports “Le Parisien”.

Advertisement

The case is all the more sensitive as it concerns a murder trial. Stephen Aarons defends on appeal Oscar Renee Sandoal, sentenced to life in prison for the murder of the mother of his children and who pleads not guilty.

To prepare the case, the lawyer provided ChatGPT with documents and asked it to summarize the proceedings of the first trial. He believed the tool would provide him with a reliable summary. But the result notably contained testimonies from people who did not exist and fabricated police statements.

The problem does not stop at the error of artificial intelligence: the lawyer transmitted these elements to the courts without checking their accuracy. It is this failure that the Supreme Court criticizes him for. She found him guilty of contempt and decided to refer his case to the disciplinary authorities.

‘I didn’t know’ AI could hallucinate

Stephen Aarons pleaded naivety. He claims he did not measure ChatGPT’s ability to produce “hallucinations”, a term used when artificial intelligence generates false information by making it appear credible.

“I am full of remorse, but I hope that the disciplinary board will take into account that this was an honest mistake,” he told Reuters.

The argument hardly convinced the magistrates. Judge C. Shannon Bacon asked him if he followed the news, noting that mistakes made by lawyers who relied on AI-invented content are now widely documented.

A phenomenon far from isolated

The case of Stephen Aarons is indeed no longer exceptional. According to the database established by law researcher Damien Charlotin, cited by “Le Parisien”, nearly 1,400 cases involving AI hallucinations in legal proceedings have been recorded in the United States, compared to at least thirteen in France. Several lawyers have already been sanctioned for submitting non-existent court decisions or erroneous legal references to the courts.

Advertisement
Jackson Avery

Jackson Avery

I’m a journalist focused on politics and everyday social issues, with a passion for clear, human-centered reporting. I began my career in local newsrooms across the Midwest, where I learned the value of listening before writing. I believe good journalism doesn’t just inform — it connects.

Leave a Comment