Donald Trump’s former lawyer and fixer, Michael Cohen, accidentally gave his lawyer phony legal citations — which were then used in a motion given to a federal judge — conjured by Google’s artificial intelligence chatbot, Google Bard.
Per The New York Times, Cohen pulled the citations when trying to aid his lawyer with a motion related to his 2018 guilty plea for campaign finance violations. Cohen spent just over a year in prison (he was released early in 2020 due to Covid-19) but remained under court supervision. The filing featuring the fake Google Bard citations was submitted in November as part of an effort to convince the court to end this supervision early because Cohen had served his time and complied with the conditions of his release.
Cohen’s extremely online goof was revealed in a set of court documents unsealed Friday, Dec. 29. Amongst them was a sworn declaration from Cohen, in which he admitted, as a non-practicing lawyer, he hadn’t kept up with “emerging trends (and related risks) in legal technology.”
As such, he continued, he was not aware that Google Bard was like Chat-GPT, and could “show citations and descriptions that looked real but actually were not.” Instead, Cohen said, he thought it was a “super-charged search engine.”
Cohen went on to claim that after he submitted the citations to the lawyer helping him, David M. Schwartz, the filing underwent “several additional rounds of edits.” According to Cohen, neither Schwartz nor his paralegal raised any issues with the citations
“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 said. “Accordingly, when I saw the citations and descriptions I had sent Mr. Schwartz quoted at length in the draft filing, I assumed that Mr. Schwartz had reviewed and verified that information and deemed it appropriate to submit to the court.”
Schwartz, for his part, submitted a declaration of his own claiming he believed the phony case summaries had been found by another lawyer working for Cohen, E. Dayna Perry. Schwartz said that because of Perry’s reputation, he “did not independently review the cases” and “failed to review what I thought was the research of another attorney.”
He added: “I never contemplated that the cases cited were ‘non-existent.’”
For those curious, the fake court cases cited in Cohen’s November motion all appear to share names with actual court cases — but the details are obviously drastically different. For instance, United States v. Figueroa-Flores was, in reality, a 2009 case about identity theft that went all the way to the Supreme Court. It was not, as the fake AI citation claimed, a case involving a defendant convicted of cocaine possession who successfully secured early termination of his court supervision by complying with the conditions of his release.
While Cohen’s AI gaffe technically has no relation to the various legal proceedings against his former boss, Donald Trump, in which he is involved, it could still be a detriment. Cohen has already testified against Trump in a New York civil case, and he’ll be a key witness in an upcoming criminal case also taking place in NYC. Trump and his legal team have long tried to cast doubts on Cohen’s reliability as a witness, and using AI — even allegedly accidentally — to cook up phony cases without doing the bare minimum to check your work probably doesn’t help too much in establishing credibility.
In the criminal case, Trump has pleaded not guilty to 34 counts of falsifying business records related to a scheme intended to cover up a hush money payment to Stormy Daniels. Cohen allegedly paid Daniels out of his own pocket, funneling the money through shell corporations, and prosecutors have alleged that Cohen confirmed with Trump that Trump “would pay him back.”
Donald Trump’s former lawyer and fixer, Michael Cohen, accidentally gave his lawyer phony legal citations — which were then used in a motion given to a federal judge — conjured by Google’s artificial intelligence chatbot, Google Bard.
Per The New York Times, Cohen pulled the citations when trying to aid his lawyer with a motion related to his 2018 guilty plea for campaign finance violations. Cohen spent just over a year in prison (he was released early in 2020 due to Covid-19) but remained under court supervision. The filing featuring the fake Google Bard citations was submitted in November as part of an effort to convince the court to end this supervision early because Cohen had served his time and complied with the conditions of his release.
Cohen’s extremely online goof was revealed in a set of court documents unsealed Friday, Dec. 29. Amongst them was a sworn declaration from Cohen, in which he admitted, as a non-practicing lawyer, he hadn’t kept up with “emerging trends (and related risks) in legal technology.”
As such, he continued, he was not aware that Google Bard was like Chat-GPT, and could “show citations and descriptions that looked real but actually were not.” Instead, Cohen said, he thought it was a “super-charged search engine.”
Cohen went on to claim that after he submitted the citations to the lawyer helping him, David M. Schwartz, the filing underwent “several additional rounds of edits.” According to Cohen, neither Schwartz nor his paralegal raised any issues with the citations
“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 said. “Accordingly, when I saw the citations and descriptions I had sent Mr. Schwartz quoted at length in the draft filing, I assumed that Mr. Schwartz had reviewed and verified that information and deemed it appropriate to submit to the court.”
Schwartz, for his part, submitted a declaration of his own claiming he believed the phony case summaries had been found by another lawyer working for Cohen, E. Dayna Perry. Schwartz said that because of Perry’s reputation, he “did not independently review the cases” and “failed to review what I thought was the research of another attorney.”
He added: “I never contemplated that the cases cited were ‘non-existent.’”
For those curious, the fake court cases cited in Cohen’s November motion all appear to share names with actual court cases — but the details are obviously drastically different. For instance, United States v. Figueroa-Flores was, in reality, a 2009 case about identity theft that went all the way to the Supreme Court. It was not, as the fake AI citation claimed, a case involving a defendant convicted of cocaine possession who successfully secured early termination of his court supervision by complying with the conditions of his release.
While Cohen’s AI gaffe technically has no relation to the various legal proceedings against his former boss, Donald Trump, in which he is involved, it could still be a detriment. Cohen has already testified against Trump in a New York civil case, and he’ll be a key witness in an upcoming criminal case also taking place in NYC. Trump and his legal team have long tried to cast doubts on Cohen’s reliability as a witness, and using AI — even allegedly accidentally — to cook up phony cases without doing the bare minimum to check your work probably doesn’t help too much in establishing credibility.
In the criminal case, Trump has pleaded not guilty to 34 counts of falsifying business records related to a scheme intended to cover up a hush money payment to Stormy Daniels. Cohen allegedly paid Daniels out of his own pocket, funneling the money through shell corporations, and prosecutors have alleged that Cohen confirmed with Trump that Trump “would pay him back.”