The Volokh Conspiracy
Mostly law professors | Sometimes contrarian | Often libertarian | Always independent
"One Would Expect … Attorneys Believe They Bring Some Level of Value to Their Clients Beyond That of a Machine"
A magistrate judge recommends a $10K sanction for a lawyer's repeated incorrect citations, and has some things to say about the pattern he has been seeing in his own court.
From Thursday's Report and Recommendation in Virgil v. Experian Info. Solutions Inc. by Judge Mark Dinsmore (S.D. Ind.):
[T]his is the third time in the last year that the Undersigned has faced a similar issue. We live in a world of rapid technological advancement. A world in which many individuals have concerns that their jobs may be replaced by increasingly educated machines.
The practice of law is not just a job, it is a profession; a profession with standards and ethical responsibilities. One of the most troubling aspects of these situations is the lack of respect for the profession, and the lack of respect by the offending attorneys for their own personal capabilities, that these situations represent.
One would expect that, when individuals choose this profession, they do so in part because they believe they have some talent for the work. One would expect that, after several years of law school, and more years of practice, those attorneys believe they bring some level of value to their clients beyond that of a machine. Yet these situations represent an abdication of those personal and professional responsibilities to those very machines, which to date have not proven themselves up to the task.
While the Undersigned has long recognized the value of the proper and efficient use of technology, my confidence in the profession and the generations of lawyers who have shaped it prevents me from believing that it can ever be replaced by a machine, no matter how advanced. However, the preservation of that profession requires ever increasing levels of diligence and vigilance from each and every attorney and judge involved in the process. Absent that, someday clients may well be better off accepting advice from a machine as opposed to a careless and inattentive attorney. That is a day the profession of law must not allow to come to pass….
Some more details on the particular transgressions in the case:
At the hearing, Mr. Policchio acknowledged that he was fully responsible for the errors in the [five] briefs that he signed and filed. Mr. Policchio was contrite and did not attempt to excuse his lapses in any way; indeed, his reaction to his own failures is best described as mortification. By way of explanation—but not as an excuse—Mr. Policchio explained that he had grossly underestimated the time that litigating this case against dozens of defendants would require and that he was overwhelmed by the workload. In addition, for monetary reasons, Mr. Policchio had terminated his access to LEXIS, which he was accustomed to using for legal research and cite-checking. The Undersigned believes Mr. Policchio's representation that he has now taken appropriate measures to ensure that no such lapses will occur in the future.
Mr. Policchio's response to the position he finds himself in is commendable. However, that does not change the fact that Mr. Policchio finds himself in this position because he failed to take the most basic of actions—checking the citations in his own briefs before filing them. As discussed below, this failure constitutes a failure to satisfy a fundamental obligation of an attorney to his client and to the Court.
{The Court notes that most of the cases cited below involve AI-generated hallucinated citations. Mr. Policchio's explanation for the errant citations in his briefs does not involve the use of AI. Rather, Mr. Policchio stated that the errors were the result of cutting and pasting from documents Mr. Policchio had saved into a case management program when working on prior cases. It is unclear to the Court how that process could have resulted in non-existent citations being created, but the distinction is irrelevant. The issue in this and the cited cases is not the use of AI to assist in legal work; it is the failure to review the validity and content of the cases cited in briefs filed with the Court.} …
The Undersigned, regrettably, has faced this issue in two other cases in the past year. In the first, monetary sanctions of $6,000 were imposed for three briefs containing hallucinated citations. {The Undersigned recommended a $15,000.00 sanction in that case: $5,000 for each of three briefs that contained non-existent citations. The District Judge reduced the sanction to $6,000, taking into account "the steps [the sanctioned attorney] has taken 'to educate himself on the responsible use of AI in legal practice' and adhere to 'the highest standards of professional conduct moving forward.'"} In the second, the Undersigned has recommended a $7,500 sanction for a brief containing hallucinated citations. The attorney's objection to that recommendation remains pending…. A multitude of other cases around the country have dealt with the same issue….
Given the distressing number of cases calling out this conduct—citing to cases in court filings without checking the accuracy of the citations—it is clear that the imposition of modest sanctions has failed to act as a deterrent. Accordingly, the Undersigned RECOMMENDS that Mr. Policchio be sanctioned $10,000.00 for his Rule 11 violations in this case. The Court finds that no lesser sanction will serve the necessary deterrent purpose…. In addition, the Undersigned REFERS the matter of Mr. Policchio's misconduct in this case to the Chief Judge pursuant to Local Rule of Disciplinary Enforcement 2(a) for consideration of any further discipline that may be appropriate. Mr. Policchio is ORDERED to provide a copy of this order to his client, … and to file a certification that he has done so within seven days of the date of this order.
Editor's Note: We invite comments and request that they be civil and on-topic. We do not moderate or assume any responsibility for comments, which are owned by the readers who post them. Comments do not represent the views of Reason.com or Reason Foundation. We reserve the right to delete any comment for any reason at any time. Comments may only be edited within 5 minutes of posting. Report abuses.
Please to post comments
This will never improve until AI stops hallucinating things.
I think the hallucinations are inherent to the technology. Inevitable.
Can't prove it, of course.
It is inherent. Partially because “hallucination” doesn’t have a precise definition beyond “saying things we deem to be factually incorrect”. But engines have no idea what we consider to be correct and no reference to some sort of truth function to judge their output against. If we look at political claims, it is evident that humans don’t have a fundamentally better truth function. AI engines predict likely words, but they do not really understand the world. As long as they do that, what we call hallucination is an inevitable outcome of the technology, and no amount of better training data or guardrails will eliminate that. It make make the problems more subtle and harder to detect, but until all their training data is in perfect agreement about an objective reality, they cannot avoid the problem.
(I am an AI researcher, by the way, who started out with machine translation, which was the genesis of current GenAI, so I’ve been on this best for a long time. Anyone who tells you that they can solve hallucination is on the wrong side of a theoretical argument that has been raging for over 1,000 years about truth correspondence and the limits of reasoning.)
A question: Is an attorney's history of misconduct, or the attorney's filings with a court explaining the "errors" admissible in a negligence lawsuit brought by said attorney on behalf client? Can a defendant in the negligence case quote the plaintiff attorney's excuses to the court in the judicial misconduct matter? Relevant to show human error rather than legal negligence, so argued the offending attorney to the court.
"One would expect that, when individuals choose this profession, they do so in part because they believe they have some talent for the work. One would expect that, after several years of law school, and more years of practice, those attorneys believe they bring some level of value to their clients beyond that of a machine."
I know I should not be shocked by this statement, but it does seem contrary to the reality of public perception of lawyers. Gallup has conducted a poll since 1976 on honesty of various professions. While not at the bottom (Members of Congress always provide a worse option) lawyers not only fare poorly they have seen a consistent decline since the poll began. By the by, while not as bad as lawyers judges don't fare much better.
To put things in perspective AI accounts for a tiny percentage of what lawyers are sanctioned for, well under 5%. The old favorites still lead the pack.
"Most Common Grounds for Lawyer Discipline (2024–2025):
Gross Neglect/Lack of Diligence: (The lawyer just didn't do the work).
Communication Failures: (Ignoring the client's calls).
Financial Mismanagement: (Messing with the escrow/trust accounts).
Criminal Convictions: (DUI, fraud, etc.)."
Conflict of Interest: (Representing two people who hate each other)."
Any of us could submit a bad brief using AI. Only those who have been admitted to the guild are allowed to do so.
Has there been any study of incidents of lawyers, paralegals, etc., submitting bad citations sans AI?
I have searched for intel on that level of resolution. There have been a little over 700 false cites using AI with over 90% in 2025. This seems to amount to between 1-2% of total sanctions by judges and state bars. My posts above lists the top five. Problem with that list is the first one seems to be a catchall that includes a lot of things.