AI Ethics for Utah Lawyers: The Rules That Actually Govern You in 2026

Here is what actually governs a Utah lawyer using AI in 2026: a published Utah Court of Appeals sanctions decision (Garner v. Kadince), a first-in-the-nation state statute (the Utah AI Policy Act), the Rules of Professional Conduct you already practice under, and ABA Formal Opinion 512 as the analytical baseline. No Utah-specific rule prohibits AI. Every Utah authority that has spoken says the same thing in different words: use it if you want — verify before it reaches a client or a court.

The map at a glance:

And one housekeeping note before the substance, because it is oddly fitting for the topic.

The Utah AI ethics opinion that doesn't exist

Search "Utah AI ethics opinion" and several AI-generated legal-guidance sites will confidently describe a Utah State Bar advisory opinion on generative AI, complete with an opinion number and issue date. Check the Utah State Bar's own ethics-opinion index and, as of July 2026, no such opinion appears — nothing on generative AI at all.

The Bar has addressed AI, but informally: a 2023 Bar Journal article, "Using ChatGPT in Our Practices: Ethical Considerations", and educational programming like the Bar's five ethical principles for generative AI. Useful reading — but not binding authority, and not the phantom opinion the content farms cite.

There's a lesson in the meta: the pages telling you how to use AI ethically appear to contain unverified AI output. Which brings us to the case where that exact failure met a real court.

Garner v. Kadince: Utah's sanctions case

In Garner v. Kadince, 2025 UT App 80, a petition filed in the Utah Court of Appeals cited multiple cases that either didn't exist or didn't say what the petition claimed. One of them — Royer v. Nelson — existed in no legal database. It existed only in ChatGPT.

The facts are the ones every small firm should study, because they're about supervision, not technology. The petition was drafted by an unlicensed law clerk who used ChatGPT. The signing attorney filed it without independently checking the citations. Opposing counsel found the fabrications — as opposing counsel always does.

The court's response: the attorneys "fell short of their gatekeeping responsibilities as members of the Utah State Bar." The sanctions ordered were pointed — pay the opposing party's attorney fees for responding to the petition, refund the client every dollar paid for it, and donate $1,000 to the legal-aid nonprofit "and Justice for All." The court was equally clear that AI itself wasn't the offense; the failure to verify was. A fabricated opinion is not "existing law" that can support any legal contention.

Utah's case fits a national pattern that has only accelerated — over a thousand tracked incidents of AI-fabricated material in filings, escalating to five-figure fines and, in 2026, the first bar suspension. The lesson has not changed once: generated citations are not research.

The Utah AI Policy Act: a real statute with a narrow bite

Utah was the first state to enact a generative-AI consumer-protection law — S.B. 149, the Artificial Intelligence Policy Act, effective May 1, 2024, and refined by 2025 amendments that narrowed its scope.

For lawyers, the operative piece is the regulated-occupation disclosure duty. Because law is a state-licensed occupation, a person providing legal services must disclose when a consumer is interacting with generative AI directly in the delivery of those services, if the interaction is high-risk — the statute's amended framing captures things like personalized legal, financial, or health advice. The disclosure comes at the start of the interaction: stated orally in a verbal exchange, in writing for a written one.

What the Act does not reach matters just as much for a small practice: a lawyer using AI internally — to research, summarize, or draft — and then reviewing and delivering the work product is not putting a consumer into an AI interaction. The statute regulates AI-facing-the-client, not AI-in-the-workflow. (An intake chatbot on your firm's website is the classic example of what does trigger it.)

The rules you already practice under

Utah's Rules of Professional Conduct needed no amendment to cover AI, and ABA Formal Opinion 512 (July 2024) — the framework Utah's informal guidance tracks — maps the terrain:

• Rule 1.1 (Competence). You must understand what a generative tool actually does — including that a general-purpose chatbot generates text rather than retrieving law. Using one without knowing that is the modern equivalent of citing a headnote you never read.
• Rule 1.6 (Confidentiality). Client information entered into a consumer AI tool may be retained, reviewed, or used for training. Before any client data touches a tool, know where the data goes, and use tools with real isolation guarantees — or strip the identifying facts.
• Rules 5.1 and 5.3 (Supervision). Garner is a supervision case. The clerk used ChatGPT; the attorney signed. Your staff's AI use is your responsibility, which means a firm policy, not a hope.
• Rule 3.3 (Candor). A fabricated citation in a filing is a false statement of law to a tribunal, however it got there.
• Billing honesty. Opinion 512 adds what Utah's Bar Journal guidance also flagged: bill the time you actually spent, not the time the task used to take.

For the national ethics picture — every state's posture, the sanctions landscape, and how the rules apply to research specifically — see our full guide to whether lawyers can use AI for legal research.

A practical checklist for a Utah small firm

1. Adopt a one-page AI policy. Who may use which tools, for what, and the non-negotiable: nothing AI-touched reaches a court or client unverified. Garner makes this a supervision requirement, not a nicety.
2. Sort your tools by retrieval vs. generation. A tool that searches real databases and links every authority is a research tool. A tool that writes fluent text from a model's memory is a drafting aid whose every legal assertion needs independent checking. Treat them differently.
3. Keep client confidences out of consumer tools. If the tool's terms don't promise your data stays out of training and out of other tenants' reach, it doesn't get client facts.
4. Verify like it's your signature — because it is. Existence, then support, then treatment, for every citation — the 10-minute pre-filing verification workflow covers all three, and our guide to checking whether a case is still good law goes deep on the last.
5. Mind the statute's edge. If clients interact with AI directly — an intake bot, an AI-drafted advice letter sent unreviewed — the Utah AI Policy Act's disclosure duty is in play. Internal, reviewed use is not.

Where this leaves you

Utah's message across all three authorities — court, legislature, and Bar — is consistent and, frankly, friendly to small firms: AI is allowed, verification is mandatory, and the lawyer remains the gatekeeper. The economics still favor adoption; the sanctions cases all involve the one step nobody should skip. (If cost is what's keeping you on a legacy platform, our breakdown of what LexisNexis actually costs pairs well with this one — and Utah's free research stack is unusually good to begin with.)

CaseRead's position on this is built into the product: answers cite only sources the system actually retrieved and linked, and anything unverifiable is flagged rather than asserted. And for the Garner failure mode specifically — AI-drafted text heading for a filing — the Hallucination Shield checks each citation it finds (up to 25 per run) for existence and support, free, no signup. It's the gatekeeping step, automated.

Frequently asked questions

Has the Utah State Bar issued a formal ethics opinion on AI? As of July 2026, no formal advisory opinion on generative AI appears in the Utah State Bar's published ethics-opinion index, despite several websites citing one. The Bar's guidance has come through informal channels — a 2023 Bar Journal article on ChatGPT and educational programming — while the binding authority comes from the Rules of Professional Conduct, the Utah AI Policy Act, and case law like Garner v. Kadince. ABA Formal Opinion 512 supplies the most detailed analytical framework.

Can Utah lawyers use ChatGPT or other AI tools? Yes. No Utah authority prohibits lawyers from using generative AI, and the Utah Court of Appeals in Garner v. Kadince acknowledged that AI can be a useful research tool. What the rules require is competence in the tool's limits (Rule 1.1), protection of client confidences (Rule 1.6), supervision of anyone using it (Rules 5.1 and 5.3), and verification of every citation and assertion before filing (Rule 3.3 and Rule 11-type obligations).

What happened in Garner v. Kadince? In Garner v. Kadince, 2025 UT App 80, a petition filed in the Utah Court of Appeals cited cases that did not exist — including one, Royer v. Nelson, found only in ChatGPT's output. The petition had been drafted by an unlicensed law clerk using AI, and the signing attorney filed it without checking the citations. The court found the attorneys fell short of their gatekeeping responsibilities and ordered them to pay the opposing party's attorney fees, refund the client's fees for the petition, and donate $1,000 to the legal-aid organization "and Justice for All."

Does the Utah AI Policy Act require lawyers to disclose AI use? In narrow circumstances. The Utah AI Policy Act (S.B. 149, effective May 1, 2024, and scaled back by 2025 amendments) requires people in state-licensed occupations — including lawyers — to disclose when a consumer is interacting with generative AI in the delivery of regulated services, if the interaction is high-risk, such as providing personalized legal advice. The disclosure must come at the start of the interaction. Using AI internally to research or draft, with a lawyer reviewing and delivering the work, is not a consumer-AI interaction under the Act.

Do I have to tell clients I use AI in my Utah practice? Not categorically. ABA Formal Opinion 512 says disclosure or informed consent is required in specific situations — for example, when client confidences will be input into a tool that may expose them, when a client asks, or when AI use is material to the representation. The Utah AI Policy Act separately requires disclosure when a consumer interacts with AI directly in a high-risk regulated service. Routine internal use of a verified research tool generally requires neither, though many firms disclose in engagement letters as a best practice.

How do Utah lawyers avoid AI citation sanctions? Verify every citation before filing — existence first, then support, then treatment. Garner v. Kadince turned on a signing attorney who did not independently check a clerk's AI-drafted petition, so the duty extends to supervising staff under Rules 5.1 and 5.3. Prefer research tools that retrieve and link real sources rather than generating citations from a model's memory, and run any AI-drafted text through a citation checker before it goes to court.