A New Wrinkle in AI Hallucination Cases: Lawyers Dinged for Failing to Detect Opponent’s Fake Citations

A new decision from the California Court of Appeals adds an intriguing dimension to the growing body of AI hallucination sanctions cases, raising the question of a lawyer’s duty to detect fabricated, AI-generated citations — not in the lawyer’s own filings, but in an opponent’s.

While the court did impose a $10,000 sanction on the attorney who filed two appellate briefs containing fake citations, it also declined to award attorneys’ fees or costs to the opposing counsel, because of counsel’s failure to report the fake citations to the court or even to detect them.

That makes this what may be the first judicial decision to touch on on whether lawyers have a duty to detect and report their opponents’ AI-generated fake citations.

Fabricated Quotations

The basic facts of the case, Noland v. Land of the Free, L.P., follow a now-familiar pattern. The appeal itself, the court said, was “unremarkable,” and would not normally have warranted publication.

But what made the decision worthy of publication, said the court, was the plaintiff’s extensive use of fake, AI-generated citations and quotations. “Although the generation of fake legal authority by AI sources has been widely commented on by federal and out-of-state courts and reported by many media sources, no California court has addressed this issue,” the court said.

The attorney, Amir Mostafavi, used ChatGPT and other AI tools to “enhance” his appellate briefs, then failed to verify the citations before filing. The court found that 21 of 23 case quotations in his opening brief were fabricated, along with many more in the reply brief. Some cases did not discuss the topics for which they were cited, and others did not exist at all.

“Nearly all of the legal quotations in plaintiff’s opening brief, and many of the quotations in plaintiff’s reply brief, are fabricated,” the court said.

Finding that both California’s Code of Civil Procedure and Rules of Court permit an appellate court to impose sanctions for filing a frivolous appeal, the court imposed a sanction of $10,000 to be paid to the court and referred the lawyer to the state bar.

“Attorney Mostafavi’s fabricated citations and erroneous statements of law have required this court to spend excessive time on this otherwise straightforward appeal to attempt to track down fabricated legal authority and then to research the issues presented without plaintiff’s assistance,” the court said.

The Twist: No Fees for Opposing Counsel

Had the court stopped there, this case might seem little different than the hundreds of cases in which lawyers’ use of AI has resulted in filings containing hallucinated citations.

But what makes Noland unique is the court’s explicit decision not to award attorneys’ fees to the opposing counsel, despite finding the appeal frivolous and despite opposing counsel’s request for such an award. The court explained:

“We decline to order sanctions payable to opposing counsel. While we have no doubt that such sanctions would be appropriate in some cases, in the present case respondents did not alert the court to the fabricated citations and appear to have become aware of the issue only when the court issued its order to show cause.”

Although the court did not elaborate beyond that statement, its reasoning raises the question: What is the role of opposing counsel in policing AI hallucinations?

Put another way: What is a lawyer’s responsibility to detect and report an opponent’s use of hallucinated citations?

The court appears to suggest that had opposing counsel spotted the fake citations and alerted the court, they might have been entitled to be awarded sanctions. Conversely, their failure to detect the fabrications made them undeserving of compensation.

The Broader Implications

Before the advent of gen AI, lawyers had been able to fairly safely rely on the assumption that cases cited by an opponent actually existed.

Of course, that did not mean that lawyers could go without checking those citations. While it was unlikely that the citation was fabricated, it could well be that the case did not stand for the cited proposition or that a quote from the case was erroneous or out of context.

So checking citations, even before gen AI, was simply good lawyering. If you are going to respond to an argument, you need to explore the foundations on which that argument was constructed.

But in the AI era, could it be that lawyers have a heightened responsibility to check their opponents’ citations? Does the responsibility now extend not just to the lawyer’s clients, but to the courts?

It is fair to say, I think, that the Noland court’s denial of attorney fees to defendants who failed to spot obvious fabrications hints at an evolving standard of professional competence.

In denying opposing counsel’s request for attorneys’ fees, the Noland court appeared to fault counsel on two counts: their failure to alert the court to the fabricated citations, and their failure to even notice them in the first place.

The Noland court’s decision seems to suggest that, at minimum, lawyers who spot AI hallucinations and alert the court may be rewarded (or at least may be eligible to be reward), while those who miss them may not be entitled to fee-shifting even when the opposing brief is ultimately sanctioned.

A Developing Standard

While Noland appears to be the first case to explicitly address opposing counsel’s role in detecting AI hallucinations, it likely won’t be the last. As AI-generated fake citations become more sophisticated and potentially harder to detect, courts will need to develop clearer standards about what level of diligence opposing counsel should exercise.

The Noland court’s terse approach of declining to award fees without defining a bright-line rule leaves the door open for other courts to provide more guidance.

Will spotting AI hallucinations become part of the expected professional competence of practicing attorneys?

My guess is that courts will increasingly expect all parties to exercise heightened vigilance in an era where fake legal authorities can be generated at the click of a button.

[Hat tip to David Kluft, assistant bar counsel in Massachusetts, who posted this case on LinkedIn.]