Jimmy Webb's narrator measures a departure in cities. By the time he gets to Phoenix she'll be rising; by the time he makes Albuquerque she'll be working; by the time he makes Oklahoma she'll be sleeping — and only then, he figures, will she believe he was serious. Phoenix is the first stop. It is the city the road runs through on the way to everywhere else in the country.
On the question of judges and artificial intelligence, Phoenix got there first. Effective January 1, 2026, Arizona became the first state in the nation to write a duty of technological competence into its code of judicial conduct. The rest of the country is still somewhere between here and Albuquerque.
This series has spent most of its attention on our side of the bench: how to use AI without getting sanctioned, what to do when the other side has leaned too hard on a chatbot, how disclaimers manufacture the illusion of free legal advice, and — most recently — what happens when a lawyer follows an AI disclosure order to the letter. This installment turns the camera around. Not because I think judges should not use AI. I think they should, and most already do. The question is what a sound rule for it should look like. And Arizona has supplied most of the answer.
She'll Be Risin’
The vehicle was Petition R-24-0052, adopted by the Arizona Supreme Court in August 2025 and effective January 1, 2026. It adds a single clause to Comment 1 of Rule 2.5 of the Arizona Code of Judicial Conduct. Judicial competence now requires “the legal knowledge, skill, thoroughness, and preparation reasonably necessary to perform a judge's responsibilities of judicial office, including the use of, and knowledge of the benefits and risks associated with, technology relevant to service as a judicial officer.”
Lawyers will recognize that language, because it started as ours. The ABA added nearly identical wording to Comment 8 of Model Rule 1.1 in 2012, and some forty states have since adopted it in one form or another. For fourteen years we have been the ones carrying an affirmative duty to keep up with the benefits and risks of the technology we use. Arizona closed the loop and applied the same expectation to the bench.
The detail I find most encouraging is who asked for it. The petition came from the Arizona Supreme Court's own Steering Committee on Artificial Intelligence and the Courts. Nobody imposed this on the judiciary; the judiciary went and got it. That same committee had already published ethical best-practices guidance for Arizona judges, and it draws the bright line where it belongs: a judge “should never use AI to reach a conclusion on the outcome of a case,” and a judge who uses these tools carries the same duties of verification and confidentiality that a lawyer does.
That is the right architecture. Prohibition was never going to hold, and pretending the tools are not already in chambers would only guarantee they get used badly and privately. Competence is the better lever, because it is a duty a judge can actually discharge, and because it puts the responsibility exactly where it has always belonged — on the person wearing the robe.
That Phone Just Keeps On Ringin'
The federal experience shows why the rule was needed. In March 2026, the Sedona Conference and the New York City Bar published a survey of the federal bench conducted by researchers at Northwestern, with a sitting federal district judge among the authors. Of the 112 judges who responded out of 502 surveyed, more than sixty percent reported using at least one AI tool in their judicial work, and 22.4 percent reported using one weekly or daily. Legal research led the list by a wide margin. Nearly half — 45.5 percent — said their court had provided no AI training at all, and roughly a quarter said their chambers operated under no policy on the subject.
We also know what that looks like when it goes wrong. In October 2025, two federal district judges — Henry Wingate in the Southern District of Mississippi and Julien Neals in the District of New Jersey — responded to an inquiry from Senator Grassley and acknowledged that AI use by their staff had produced orders containing serious errors: misquoted state law, individuals identified who were not parties to the case, and quotations attributed to defendants that did not exist. Both judges withdrew the orders and tightened their procedures. Judge Wingate now requires independent second review of every draft with the cited cases printed and attached; Judge Neals adopted a written policy barring clerks and interns from using AI to draft opinions or orders at all.
I want to be careful about the moral of that story, because it is not that judges are careless. It is the same lesson I wrote about in the context of law firms: those were supervision failures. In both cases a staff member used the tool, the draft read cleanly, and nobody caught the problem before it went out under a signature. A chambers is a small law office. It has a principal, junior people who draft, and now a machine that writes more fluently than either while being wrong more convincingly. Every institution using these tools is learning that at the same time, and courts got no head start on the rest of us.
The Note I Left Hangin' on Her Door
Here is what Arizona has not yet supplied, and neither has anyone else: competence is not transparency. A duty to understand a tool says nothing about whether the parties are ever told the tool was used.
The federal judiciary's interim AI guidance, issued in 2025 and never released publicly in full, gets the accountability piece right. Users remain “accountable for all work performed with the assistance of AI,” and the guidance cautions against delegating core judicial functions, including decision-making and adjudication. On the question of telling anyone, though, it goes only this far: judges should “consider whether AI use should be disclosed.” Arizona's guidance likewise stops short of requiring it. So we have arrived at a competence duty with no transparency mechanism attached to it.
What would one look like? The line I would draw runs between information and judgment, and it is not difficult to state. Using these tools to find authority, summarize a record, check a citation, or tighten prose is research and drafting assistance; nobody needs to be told, from either chair. Using them to weigh credibility, resolve a disputed fact, or select an outcome is not assistance at all — it is the delegation that Arizona's committee and the federal guidance both already prohibit, and no amount of disclosure would cure it. What lives between those two poles is a narrow band that deserves a note: one sentence in the order saying that AI was used in its preparation and that the court has verified the authorities cited. The song's entire device is a man who leaves a note on the door. It costs him nothing, and it is the only reason the story is fair to the person who reads it.
I will be accused of projection here, and the accusation is at least partly fair, so let me be exact about what I am asking. I am not arguing for less disclosure by lawyers. Some standing orders now require counsel to certify every point at which AI touched a filing, and I have had my fun with the more expansive versions of that requirement in an earlier post. My argument is for one standard instead of two: the same line between information and judgment, the same duty to verify, and the same accountability attached to the signature at the bottom of the page — whichever side of the bench that signature comes from. If the rule I have just proposed is too burdensome for a court to follow, it is too burdensome to impose on counsel. I am content to be held to it either way.
By the Time I Make Albuquerque
There is a practical consequence for those of us who appear in these courts, and it has nothing to do with disclosure. Rule 2.5 places the competence duty on the judge. But nearly half of the judges in that survey said no one had trained them, and a rule does not arrive with a syllabus. A duty to understand the benefits and risks of a technology, landing without institutional teaching behind it, gets discharged through the record — which means that in most cases it gets discharged by counsel.
That is a larger strategic fact than it first appears. In a case that turns on how a model was built, what data trained it, or whether an output can be trusted, the party that credibly teaches the court the technology has enormous influence over how the court rules on the technology. That job belongs to the expert you retain and, sometimes, to the co-counsel you bring in. Which is where this series goes next.
This post is part of an ongoing series on how AI is reshaping litigation and law-firm practice. Earlier installments addressed AI, sanctions, and the new standard of competence; what to do when the other side trusts a chatbot; the illusion of free legal advice in AI disclaimers and terms of use; and a rather literal exercise in complying with an AI disclosure order.
A Word About Silver Cain
Silver Cain PLC represents businesses in complex commercial and real estate litigation in Arizona and beyond. When Rebecca Cain and I founded the firm, we built it around direct partner involvement and senior trial-level judgment — including the judgment to say plainly what we think a good rule looks like, whichever direction it happens to point. If the questions in this post are relevant to your business, or to the firms you retain, we are glad to have that conversation.
Silver Cain, PLC is an Arizona based boutique law firm with recognized excellence in commercial and real estate related disputes. The firm represents businesses, investors, and professionals in sophisticated legal matters throughout the United States, providing nuanced strategies and high-caliber advocacy in both state and federal courts, arbitrations and mediations. Our attorneys work closely with business owners, investors, and individuals to protect their interests and find the best path forward, making Silver Cain, PLC a trusted resource for clients and fellow professionals alike.
