Be Careful What You Tell AI
Typing a question into ChatGPT, Claude, or another AI tool may feel like asking a lawyer privately. It isn’t necessarily the same thing.
A client in a contract dispute might upload the contract and ask AI to interpret a clause. A divorce client might paste text messages with a spouse into a chatbot. A business owner might describe a dispute with a former employee and ask AI to spot potential claims.
Each of these creates a digital record of the exchange — and, as recent cases show, that record may or may not be protected from disclosure.
The cases so far are mixed.
In United States v. Heppner (S.D.N.Y. Feb. 17, 2026), a criminal defendant used Claude to help build defense strategies and shared the output with his attorneys. The court held that the AI output must be turned over to the other party because neither attorney-client privilege nor work-product protection applied: the AI conversations weren’t communications with an attorney, weren’t confidential enough for privilege, and reflected the defendant’s own work rather than work directed by counsel.
In Warner v. Gilbarco, Inc. (E.D. Mich. Feb. 10, 2026), however, a pro se plaintiff used ChatGPT in her litigation. The court denied the defendants’ request to discover her AI use, finding it protected by the work-product doctrine. The court noted that generative AI tools are “tools, not persons,” and that using one to draft doesn’t waive that protection.
Takeaway: there’s no blanket rule that AI conversations are always discoverable — or always protected. Courts are still applying ordinary privilege, work-product, and relevance principles to this new context.
The safest approach: don’t put anything into an AI system that you’d be uncomfortable handing to the opposing party. Asking “What is the difference between mediation and litigation?” is very different from uploading court filings, describing privileged conversations with your attorney, naming witnesses, or detailing settlement strategy.
What This Means Under Utah Law
Utah’s discovery rules (Utah R. Civ. P. 26) already cover electronically stored information, allowing discovery of relevant, nonprivileged material that’s proportional to the case. AI conversations are simply a new form of that same category — so whether a given prompt or output must be produced depends on the usual factors: relevance, privilege, work product, and proportionality.
The law here will keep developing. Until it is fully developed, treat AI the way you’d treat any other technology that creates a record.
Millar Legal’s Approach to AI
Utah lawyers aren’t barred from using AI — in fact, Utah’s Rules of Professional Conduct treat staying current on technology as part of competent representation. Rule 1.1’s comment specifically ties competence to understanding the benefits and risks of relevant technology.
Millar Legal takes that seriously and takes client confidentiality just as seriously (Utah R. Prof’l Conduct 1.6(a)). When we use AI for drafting or research, we use only enterprise-grade tools built for data security, and every output is reviewed by an attorney for accuracy before it’s used.
Bottom Line
An AI chatbot is not your lawyer. A conversation with one may not carry the same legal protections as a conversation with your attorney — so think before you click “send.”
**We help people protect and enforce their rights in divorce, custody, paternity, enforcement, and modification proceedings. Please let me know if we can help someone you know.