Trust but Verify

Many attribute the phrase "trust but verify" to Ronald Reagan. It was undoubtedly stressed by him and his team during the 1980s arms negotiations that brought some respite to the nuclear threats of MAD—Mutually Assured Destruction. If you are too young to remember the Cold War, the U.S.S.R., the Berlin Wall, and such, bless you. Despite that clear example of the phrase, some attribute it to Winston Churchill and others.

The reader will likely recall some posts here that focused on artificial intelligence (AI). I have penned a couple over the years. I am pleased to have been invited to speak on AI at various conferences and have been studying it, robotics, and their impacts on the world of work for over a decade. Nonetheless, there are more prolific AI writers, one of whom I married long ago, in a galaxy far, far away.

As she and I recently discussed the AI threat to legal systems, I mentioned "the judicial example" that I found so disappointing. Though she agreed, our conversation soon revealed that we were each citing the same problem, a judicial disappointment, but we clearly were not on the same page factually. I was focused on the new(er) judge in Georgia whose order was vacated by an appellate court for being fiction. See Better Look That Up (July 2025).

However, she was referencing a more recent example(s) of judicial criticism cited recently by Reuters. These instances involve federal judges "in Mississippi and New Jersey." Judge Henry Wingate of Mississippi was appointed in 1985 by Ronald Reagan. Judge Julian Xavier Neals was appointed by Joseph Biden in 2021 and has been on the bench for four years.

According to Reuters, Judge Wingate entered an order "in a civil rights lawsuit." Lawyers then brought to his attention that the order "contained 'incorrect plaintiffs and defendants' and included allegations that were not in the complaint." The lawyers contend that "the decision contained a number of inaccuracies."

Judge Neals "withdrew a ruling he issued in a securities lawsuit." That order allegedly included "factual errors and ... quotes that the lawyers said were not in the cited cases." Allegedly, "the decision in three places misstated the outcomes of other lawsuits that the ruling cited." The situation was explained by one observer, noting:

1. "Research produced using artificial intelligence was included in a draft decision." And that
2. "was inadvertently placed on the public docket before a review process," and
3. "The court's chambers has a strict policy against the unauthorized use of AI to support opinions."

Reuters notes that in neither case did the lawyers make allegations about "how the apparent errors were included in ... decisions." They "did not mention AI or other possible causes for the alleged inaccuracies." 

There is a certain amount of reticence involved in correcting the judge. Every lawyer knows that, feels that, and reacts accordingly. I, for one, appreciate lawyers pointing out mistakes and errors. Rehearing is far preferable than an order reversed by an appellate court for fiction. And, rehearing is far more rapid than the time required for appellate review. 

In Better Look That Up, I noted that judges should be able to count on four things. The fourth is likely the most important (we are only human). But the corollary is as compelling. Lawyers should be able to count on judges:

1. Not to rely on what a pleading says that a statute or a case says or stands for.
2. To look at the actual statute and published opinions before ruling.
3. To be open to and introspective about motions for rehearing that cite concerns about misapprehensions or even hallucinations.

Reuters cites a law school professor who concluded that there is potential for error. His caution was clear: "These errors are going to come out if you don't pay very careful attention to the output." Errors are going to come out regardless; they will persist if we are not persistently attentive. 

AI is here. No retreat path is either practical or debatable. I am drawn in that regard to lyricist Jenna Raine:

Okay! It's not the end of the world

There's always light where it burns

Just remember the words, and it goes like this

It is what it is, what it is, what it is, what it is

It is what it is, what it is, what it is, what it is

And there is the critical point to which we return time and again. Certainly, lawyers are obligated to tell the truth. When a lawyer says "it is what it is," other lawyers and judges should be able to rely on that. There are ethical rules of candor, and that should be the end of it. Lawyers speak the truth.

That said, there will be mistakes made—"it is what it is."

Lawyers will use AI—"it is what it is."

Judges will use AI—"it is what it is."

"strict policies against ...use of AI" will be broken—"it is what it is."

And, in truth, "It's not the end of the world." It is potentially corrosive. It is likely distracting. It is going to be periodically embarrassing for the lazy or inattentive. I am now also confident that it will be pernicious, persistent, and at times perplexing.

The cure? It is simply "Trust but verify." Whether it is written in a pleading or an order, it should be read critically and carefully. If there is pause, concern, or question that should be raised by parties, attorneys, clerks, and judges. Those case citations should be checked. Those factual findings or allegations should be checked. Verify, verify, verify, just like Churchill's thoughts on surrender: "never, never, never."

It is not that these efforts will render perfection. It is not that rules ("strict policies against") will preclude errors. It is simply that computers allow us to make mistakes at an epic pace. They empower and enable in ways that are the truth of our modern day. Humans also make mistakes. No one can change these absolute truths. "it is what it is."

The only hope we have for minimizing their errors is in our human, patient, and careful verification. This means reviewing what is generated by staff, by lawyers, by judges, and by AI. Review and verify.