Law is one of the professions that embraced the use of generative AI even before the ChatGPT boom arrived. We have known cases of Lawyers sanctioned for including false quotesbut that hasn’t deterred others from doing the same. The latest case we heard about takes it to a new level: both the prosecution and the defense used AI and the judge said enough.
what has happened. It has happened in Mississippi, United States. The case in question involved a lawyer, named Tom Withers, who was seeking unpaid legal fees from the city of Aberdeen. According to what they say in 404mediaeverything was going well until the judge noticed that there was something strange in the briefs presented by the lawyers of both parties: both cited cases that did not exist. Come on, they had used AI and had not even stopped to check for inaccuracies.
Consequences. The lawyers acknowledged having used AI without verifying the results. Judge Sharion Aycock not only decided to suspend the trial, she also disqualified two of the four lawyers involved from appearing before her court for two years. The others were not spared and all received a fine that ranged between $1,000 and $3,500. Withers, the plaintiff attorney, was not representing himself, so he has escaped sanction.
Why it is important. This case represents an increasingly common trend in the legal sector. Already There are many lawyers caught using AIbut also this time it takes it to a new level because all parties have fallen into temptation. Just as the lawyer said Rob Freund in X“there were two clients who were basically paying for ChatGPT to argue with itself.”
In Spain too. Dozens of cases have already been recorded in the United States, but it is not a trend exclusive to there. Just a few days ago, The Superior Court of Justice of Galicia initiated an investigation against a lawyer for “procedural bad faith.” Apparently, the appeal he presented contained no less than “24 false jurisprudential citations.” He has not been the only one, at the beginning of the year a lawyer was fined 420 euros in the Canary Islands to cite 48 false sentences.
The hallucinations. They remain the Achilles Heel of LLMs and they have a lot to do with how they are trained: the models always prioritize giving an answer, even if it is inaccurate. There are ways to minimize hallucinationsbut at this point we all know these weaknesses, which is why these cases are so striking. The serious thing is not that lawyers are using AI, but that some do so without any type of verification. Of course, it is unknown what tools they used, but they exist specific AI tools for lawyers that promise greater reliability.
Image | Xataka with Gemini
GIPHY App Key not set. Please check settings