Donald Trump's former lawyer Michael Cohen says he mistakenly used bogus legal citations from Google Bard in a motion seeking an early end to court supervision
“The revelation could have serious implications for the Manhattan criminal case against #Trump, in which #MichaelCohen is expected to serve as the star witness. The former president's lawyers have long attacked Mr. Cohen as a serial fabulist; now, they will have a brand-new example.” #legal … Bluesky: Paul Konstant / @progcpl.bsky.social : “mistakenly” 🤣🤣🤣 [embedded post] X: @scot1andt : Not sure what's worse. Cohen using AI for his legal research or his attorney not checking his citations. ouch. Sarah Frier / @sarahfrier : “He said he had thought of Bard as a ‘super-charged search engine’ and not a service that would generate real-looking but phony legal cases.” ... Cohen should know better, but when it's at the top of Google search, this is what a lot of people will think. https://www.bloomberg.com/... Aaron Levie / @levie : This story was obviously bound to happen [image] Jeffrey Evan Gold / @jeffgoldesq : Michael Cohen was never a good lawyer. His own lawyer knew that & should have checked Cohen's AI fueled research featuring nonexistent cases. I suspect the judge will seriously reprimand but not punish here. Yet I could see a fine for not checking cites. https://www.nytimes.com/... Andrew Egger / @eggerdc : With enemies like these, who needs friends? https://www.nytimes.com/... [image] Chad West / @chad_westreal : @Techno_Fog Cohen is not a “non-lawyer” — he graduated law school and even practiced law — he's a “former attorney.” He should know better even if he's on the bench professionally. [image] @techno_fog : A few weeks back, a federal judge noticed attorneys for Michael Cohen were using fake cases - and demanded an explanation. It turns out Cohen was using AI to help write his own motions - and was apparently misleading his lawyer about the source of those cases. [image] Forums: Msmash / Slashdot : Michael Cohen Used AI To Feed Lawyer Bogus Cases
Context & Ripple Effects
This is a second high-profile legal-filing episode involving generative AI: a New York lawyer had already faced sanctions over a ChatGPT-produced brief containing fabricated authorities. Cohen’s account extends the problem from drafted arguments to legal research treated as a search tool.
It matters beyond the filing because Cohen is expected to be a key witness in the Manhattan case, and the reported citation error gives opponents a concrete new line of attack on his reliability.
First-order effects
- Cohen’s motion and his explanation for its citations face immediate credibility pressure; the issue also supplies Trump’s lawyers with material for challenging Cohen’s trustworthiness.
- The episode exposes a direct verification failure: Bard returned plausible-looking authorities, but neither their existence nor their applicability was adequately checked before filing.
Second-order effects
- Law firms and individual lawyers using generative AI for research have stronger reason to require citation-level review, particularly after the earlier sanctions case over fabricated ChatGPT citations.
- AI providers marketed or perceived as search-like tools face greater scrutiny over whether users understand that fluent outputs are not verified legal sources.
Third-order effects
- If such cases continue, legal AI adoption is likely to be organized around provenance, human review, and auditable research workflows rather than unrestricted chatbot use.
- Courts and professional bodies may increasingly treat AI-generated authority errors as a responsibility-of-counsel issue, regardless of which model produced them; later UK warnings over AI-generated fake legal material point in that direction.
The trend: Generative AI is moving into professional workflows faster than verification practices, making source reliability and accountability central adoption constraints.