Phony cases continue to proliferate across the docket. This recent explosion stems from the advent of artificial intelligence tools, with over 700 instances of embarrassing hallucinations working their way into filings so far. The problem will inevitably get worse since these AI tools are eager to provide users with whatever answer they desire, even if it’s wholly made-up garbage. That’s not entirely the fault of the AI. A non-savvy user is more likely to prompt the tool in ways that incentivize the algorithms to produce results to match the user’s request. Ask a large language model for key landlord-tenant citations and it will — often — do a decent job. Ask it to provide case citations for the proposition that my bonkers argument is actually correct, and it has a much higher chance of going off the rails.
Some tools have more robust safeguards than others, but, at the end of the day, a large language model wants to give the user what it wants. That’s trouble if the user makes the wrong ask and isn’t careful about checking the work of their semi-random word generator.
While lawyers keep screwing this up, the pro se litigant presents a vector for hallucinatory infection. They’re already up against it with a system they don’t fully understand and AI provides easy, seemingly right answers. If AI is mansplaining-as-a-service — exceedingly confident, regardless of accuracy — then its most trusting victims will be people just trying to figure out how to enforce their rights.
And it’s a problem bound to get worse because AI is cheap and lawyers are expensive.
That said, once the courts warn a litigant to stop using AI, that should be the end of it. One litigant, however, went the other direction and claimed the court’s warning proved its bias against his case.
This argument fared… poorly.
Finally, Plaintiffs [sic] objects to the Magistrate Judge’s “criticism” of his use of artificial intelligence to cite to non-existent case law and errors in other citations. Id. at 3 (citing Non-Final R&R at 2-4). Notwithstanding that a review of Plaintiffs “citations” proves the Magistrate Judge’s point, the warning given by the Magistrate Judge with respect to Plaintiffs future filings had no impact on the full analysis conducted by the Magistrate Judge on each of Plaintiff s claims.
Rob Freund (who flagged the opinion on the platform formerly known as Twitter) offered the friendly advice “If a judge calls you out for citing “non-existent case law,” filing an objection in response is probably not the play.”
The plaintiff’s specific objection was that the magistrate judge’s warning was unclear:
No Specific Misquotes; Opposing Misstatements, Overlooked [Doc. 65 p. 2] R&R Error: Vague “AI-generated, incorrect laws” claim [citing Doc. 12-1] without examples.
The problem with this claim is that there were, in fact, multiple specific examples of false citations. They were laid out by opposing counsel in its motion. Most notably the response to plaintiff’s citing “Solomon v. Norwest Corp., 546 S.E.2d 330 (Ga. 2001),” prompting opposing counsel to write:
The citation of “546 S.E.2d 330” is actually for the case Nunley v. Nunley, 248 Ga. App. 208, 546 S.E.2d 330 (2001), involving a hen farm partnership.
Talk about laying an egg.
(Full opinion on the next page…)
Joe Patrice is a senior editor at Above the Law and co-host of Thinking Like A Lawyer. Feel free to email any tips, questions, or comments. Follow him on Twitter or Bluesky if you’re interested in law, politics, and a healthy dose of college sports news. Joe also serves as a Managing Director at RPN Executive Search.
The post Warning Party To Stop Citing Fake AI Cases Is Not, In Fact, Bias appeared first on Above the Law.