In August 2025, I noted the hubbub surrounding two federal judges in Trust but Verify (August 2025). These instances involve federal judges "in Mississippi and New Jersey." Judge Henry Wingate (78) of Mississippi was appointed in 1985 by Ronald Reagan. Judge Julian Xavier Neals (60) was appointed by Joseph Biden in 2021 and has been on the bench for four years. What do the two share? Well, neither one is part of the tech-savy generation "alpha" or "z." Judge Neals is Generation X and Judge Wingate is a Baby Boomer, but an early Boomer. They each made the news for artificial intelligence and its implications.
Repetition of one critical point is necessary in every post—you cannot trust artificial intelligence. It is a boon, a powerful tool, and it should be used. As time passes, it will become ubiquitous and will be more commonly used and relied upon. I was shocked on a recent AI presentation when the sponsor deployed an impromptu poll, "Are you using AI?" and the results did not support universal commitment to the technology. That will change. That said, more than half of the poll respondents admitted to currently using it.
More recently, reports noted that the U.S. Senate Judiciary Committee inquired about the circumstances surrounding Judges Neals and Wingate. Both federal judges explained that the "error-ridden" "decisions ... did not go through their chambers' typical review processes before they were issued." Simply stated, there was trust there, but insufficient verify.
Judge Neals explained that "a law school intern used OpenAI's ChatGPT for research without authorization or disclosure." Judge Wingate similarly placed responsibility on "a law clerk ... (who) used Perplexity 'as a foundational drafting assistant.'" There is no explanation in the news report regarding why an intern or a law clerk was releasing decisions in any event, or how that would occur without the review and agreement of the judge.
Some will believe that all case decisions are made by judges. That should be a fair and reasonable assumption. Judges are elected and appointed to make difficult decisions. The news article seems to suggest that in at least these two courts, decisions are made and published by interns and clerks. Delegation is one thing, but responsibility is another. No decision is issued by the OJCC except by a judge. Perhaps that is true more broadly (most tribunals), or perhaps our judges are somehow an exception?
Some will take my thoughts as criticism. Certainly, others have been more direct in their thoughts regarding the Judges, their decisions, and their deflection onto staff. See Admit the Obvious, Above the Law. Nonetheless, my thoughts are not a criticism of judges or courts. That said, lessons can be learned by all whenever any of us makes an error.
Federal judges are governed by the Code of Conduct for United States Judges. This is arranged in canons, as are the various state codes that have been periodically discussed here. Canon 2(A) reads:
(A) Respect for Law. A judge should respect and comply with the law and should act at all times in a manner that promotes public confidence in the integrity and impartiality of the judiciary.
Canon 3(a)(5)
(5) A judge should dispose promptly of the business of the court.
There is no provision that suggests that law clerks and interns should dispose of the business of the court.
There is the potential that decisions being published by clerks and interns could impact "public confidence" in courts and their decisions. As has been frequently emphasized, these Codes preclude not only impropriety but also the "appearance of impropriety." The commentary to Canon 2 includes: "A judge must avoid all impropriety and appearance of impropriety."
Judge Wingate concluded that the failure in his chambers "was a lapse in human oversight." That is, a failure in the "verify" element of the Trust but Verify suggestion. Judge "Neals said his chambers has since created a written AI policy and enhanced its review process." Potentially, this means interns in that chamber are no longer issuing decisions.
Judge Wingate, caught in the wake of this mistake, reportedly deleted "the order (that) was riddled with mistakes" from the case docket, published a new order, and backdated that new order to the date of the original hallucinated one, according to News from the States. Does deleting the error make it go away?
Each judge asserts that they have instituted new processes to assure the errors will not be repeated. According to the Magnolia Tribune, Judge Wingate's solution is simple. Now, he assures,
all draft opinions, orders, and memorandum decisions undergo a mandatory, independent review by a second law clerk before submission to me.
Does this mean that the previously published decisions were, in fact, submitted to Judge Wingate before publication? If so, responsibility for the hallucinations rests with the judge.
Would any tribunal, anywhere, excuse a lawyer who signs a document full of hallucinations on the premise that "my paralegal" or "my associate" wrote this and I just signed it? I am dubious regarding the efficacy of such an excuse.
What does "verify" mean? There has been seeming confusion about that in the world of workers' compensation. I have striven to deliver suggestions in A Tool Kit (October 2025) and Simple Steps (October 2025). There are various methods that will help you to ensure the absence of hallucinations in your filings.
What does "verify" not mean? Well, that you had one paralegal, clerk, or intern check the work of another paralegal, clerk, or intern is not likely compatible with your personal professional responsibility.
Similarly, knowing that AI hallucinates, would you ask a second AI (that hallucinates) to verify the output of your primary AI? Doubtful anyone would view that as a rational process.
No, in the end, the signatory (judge, lawyer, engineer, doctor) is responsible. They can perhaps delegate some of the work, but it is their job to do the verification and verifying. In these two instances, as in all occasions where some staff, associate, or partner does work, it all comes down to the person who signed the final document. The buck must stop there.
The Nancy Reagan method is most effective ("just say 'no'"), meaning just don't use AI. I highly recommend this one regarding legal research, writing, and pleading. Use AI to check grammar, spelling, or similar. Do not trust it to do your work for you.
Second, there is the Ronald Reagan method, "Trust but Verify," which means the person who is hired to do the work (judge, lawyer, engineer, etc.) may delegate but must remain personally responsible.
Finally, there is the laissez-faire method in which this new and proven unreliable technology is engaged and trusted without question, as are clerks, interns, paralegals, and more. I cannot recommend that last one, but the decision is up to you.
The person signing the document makes the decision. Appropriately so. It is their name that will be on the line. See Do You Care About Reputation (June 2023). The lawyer, doctor, engineer, judge, or other person who signs the document is responsible. This will remain whether you claim your dog ate your homework or your paralegal, intern, or clerk made mistakes in judgment or process.
The paralegal, clerk, intern, associate, partner, etc., will not be the one whose reputation is impacted, impugned, or ruined. Review the Reuters article. What are the names of the clerk and intern in these chambers? They are not identified. Various searches for their identities have come up short. In this instance, it appears that the judges instead will be the ones impacted. That said, the names of clerks and interns may eventually surface.
It will persistently be the lawyer, doctor, engineer, or judge who is embarrassed or liable for errors. The implications may be an "error-ridden decision" published by an intern, a hallucination-filled pleading filed by a paralegal, or a set of structural plans completed and mailed by an engineering intern. The result will likely be similar in each instance: embarrassment, ridicule, or worse.
Will the world remember the name (or even ever know the name) of the paralegal, clerk, or intern? Unlikely. Various queries have failed to identify either the clerk or intern by name here. Will those called upon to review the results take pity on, or commiserate with, the judge, engineer, or lawyer who seeks to cast blame on their staff, team, or contributors? As unlikely. In the end, the world will look to the professional to be right, diligent, and thoughtful in the process for which they are paid.
It is entirely possible that the world will hold those people to a standard that it would not hold itself. Will judges who fail to supervise be empathetic towards lawyers appearing before them who have similarly failed to supervise their staff?
Will it take a bridge collapsing through unchecked math? Will it take a lawsuit dismissal? Will it take a lawyer disbarred, a judge impeached, or an engineer stripped of a license? What must happen before professionals take the challenges and shortcomings of AI-produced material seriously and check the work on which they are placing their name?
It is encouraging that each of these judges has implemented plans to prevent these errors from recurring. Every lawyer, engineer, accountant, doctor, etc. should take heed of the potential AI embarrassment to themself, their office, and their professions. A sound AI policy is critical. And no, having some other clerk or AI check someone else's (or some other AI's) work is not a policy.
You can delegate tasks, but not responsibilities. When you get caught, deleting it will not make it go away. Deleting it will just make for more questions.
