Brendan McDermid/Reuters
- Michael Cohen was looking for cases to back up his legal claim, so he turned to Google's AI chatbot, Bard.
- But the cases were bogus.
- Cohen's lawyer never checked that the cases actually existed before mentioning them in court.
The latest victim of an AI screw-up? Donald Trump's former fixer Michael Cohen and his lawyer.
Cohen admitted in a sworn statement in a Manhattan federal court case that he used Google Bard, a generative AI chatbot similar to ChatGPT, to find legal cases backing up his arguments for why he should be let loose early from his supervised release.
But Cohen didn't realize a key AI pitfall: sometimes, it just makes stuff up.
Cohen wrote that he misunderstood Google Bard as a search engine, not a generative AI service like ChatGPT, and that he trusted his lawyer to verify the cases.
"I understood it to be a super charged search engine and had repeatedly used it in other contexts to (successfully) find accurate information online," Cohen wrote. "I did not know that Google Bard could generate non-existent cases."
Cohen fed Bard's hallucinated results to his lawyer at the time, David Schwartz, who included three of them in his November 29 filing without checking that the cases were actually legit, according to the court papers.
The Friday legal filing was first reported by Inner City Press' Matthew Russell Lee.
Cohen, in the court documents, deflects the blame on his lawyer for not double-checking what he sent.
"It did not occur to me then— and remains surprising to me now—that Mr. Schwartz would drop the cases into his submission wholesale without even confirming that they existed," Cohen wrote in the filing.
Cohen pleaded guilty in 2018 to campaign finance violations and served time in prison before he was placed on supervised release.
The court filings had argued that he complied with all the terms of release and that his supervision should now end.
US District Judge Jesse Furman, who's overseeing the case, asked Schwartz "to explain why he should not be sanctioned for citing cases that appear not to exist," leading to Cohen explaining his side of the story.
Schwartz, in his own court filing, said he included the hallucinated cases as citations because he wrongly understood them to come from another one of Cohen's attorneys, Danya Perry, rather than Cohen himself. Perry is still working to try to end Cohen's supervised release and says he should not suffer for his lawyer's alleged misstep.
The incident could have consequences for one of the criminal cases against former President Donald Trump.
Cohen is expected to be a star witness in the Manhattan District Attorney's case against Trump, alleging he falsified business documents to cover up hush money payments to Stormy Daniels ahead of the 2016 election.
Trump and his attorneys have long argued that Cohen could not be trusted, given his criminal history, which also resulted in him losing his own legal license. With the AI snafu, they may have yet another example to bring in front of a jury to try to discredit him.