AI Is Already in Your Practice. Arizona Has Now Told You What That Requires
By Kent Berk on August 19th, 2026 in BLOG, Artificial Intelligence, Estate Litigation
Takeaways from the MCBA’s “Mind, Machine & Practice” forum
Today, I moderated the second session of the Maricopa County Bar Association’s quarterly forum series, Mind, Machine & Practice. I was joined by the Hon. David B. Gass of the Arizona Court of Appeals, Division One, and the Hon. Lisa VandenBerg, Arizona’s Presiding Disciplinary Judge. Our subject was artificial intelligence in legal practice, and the hour was eligible for ethics CLE credit.
What follows are my own takeaways. Nothing here is attributed to either judge, and nothing here is legal advice.
Arizona has moved from talking about AI to regulating it
Three developments in the last twelve months framed the conversation.
Effective January 1, 2026, Arizona became the first state in the country to write technology competence into its Code of Judicial Conduct. Rule 2.5, comment 1 now provides that judicial competence includes the use and knowledge of the risks and benefits of technology. That duty now mirrors what ER 1.1 has long required of lawyers.
In July 2026, our Court of Appeals decided Dineen-Shibata v. Kotchka, 1 CA-CV 25-0606-PB. A litigant submitted an appellate brief in which six of eight citations were deficient and two were to cases that simply do not exist. The court held that filing fabricated, AI-generated authorities is sanctionable regardless of intent, and refused a good-faith defense in terms worth quoting: “Submission of hallucinated case citations—intentional or unintentional—is unjustifiable. To accept a defense of lack of ill intent would legitimize reckless indifference to the truth and accuracy of court filings and tolerate some level of misrepresentation. This we will not do..”
And a petition is pending before the Arizona Supreme Court, proposed Rule 135, that would go considerably further on the judicial side. It would provide that “[t]hrough December 31, 2029, no judicial officer may use Generative AI in core judicial work,” meaning the drafting of orders, rulings, decisions, and opinions, and would require disclosure where AI did assist. There have been numerous comments and objections to this rule.
So the question is no longer whether AI belongs in our practice. It is here. The question is what competence, confidentiality, candor, and supervision now require of us.
The lawyer is still the lawyer
The theme that ran through the entire hour was responsibility. AI is a tool that requires supervision and verification. It is not a substitute for professional judgment, and it does not shift accountability for what goes out the door under your signature. Kotchka makes the consequence of skipping that step concrete. Good intentions are not a defense.
Self-represented litigants are changing the math for the courts
One thread I had not fully appreciated was the burden on the bench. Self-represented litigants increasingly use AI to generate pleadings, and courts are absorbing the cost of evaluating whether the cited authorities are real. That is time that comes out of every other case on the calendar.
The discussion included practical responses to serial AI-generated filings by pro se parties: meeting and conferring with the litigant, requesting a status conference, and, where the pattern warrants it, seeking vexatious litigant relief.
There is a real tension here, and we did not resolve it. Kotchka itself involved a self-represented party in a probate appeal. A no-intent sanctions rule falls hardest on people who reached for AI because they could not afford counsel. Whether that deters fabrication or deters access is a question the profession will be working through for some time.
Avoidance is not a strategy
The panel was clear that ignoring how prevalent AI has become is not a responsible posture. The better response is education and deliberate adoption, with the firm balancing privilege, confidentiality, accuracy, and ethical obligations against real efficiency gains.
That means guardrails: a written firm AI policy, training, detailed instructions and parameters inside the tools themselves, and a review step that is actually performed rather than assumed. It also means knowing what a given tool does with what you put into it, before client confidences go in.
At Berk Law Group, we use AI on the administrative side for things like forms, employee handbook content, and website applications, and on the legal side for research support, drafting assistance, and workflow. Every output gets reviewed. That is the whole point.
The rules were already there
Here is what I keep coming back to. Not one of the Arizona ethical rules implicated by AI is new. ER 1.1 competence, ER 1.3 diligence, ER 1.6 confidentiality, ER 1.4 communication, ER 1.5 fees, ER 3.3 candor to the tribunal, ER 5.1 and 5.3 supervision, ER 8.4(c) misconduct. Every one of them predates generative AI by decades.
What changed is the speed. A single tool, used carelessly, can now implicate five of those rules in one filing, in about ninety seconds. The obligations did not get harder. The failure mode got faster.
Coming up
The next installment of Mind, Machine & Practice is Thursday, November 12, 2026. Information is available through the Maricopa County Bar Association.
Kent S. Berk is the managing attorney of Berk Law Group, P.C., where he handles probate, trust, estate, and fiduciary litigation throughout Arizona. This article is provided for general informational purposes, is not legal advice, and does not create an attorney-client relationship.
