Utah ties AI disclosure duties to regulated occupations

The persona

No single persona is profiled here. The entry reads Utah's enacted AI Policy Act for what it actually requires of a business whose customer-facing product is, or resembles, an interactive virtual persona reaching Utah consumers. The bill status page identifies it as S.B. 149, "Artificial Intelligence Amendments," sponsored by Senator Kirk Cullimore and floor-sponsored by Representative Jefferson Moss, signed by the governor on 13 March 2024.

What the documents establish

The enrolled bill text takes effect 1 May 2024 and enacts Utah Code 13-2-12, defining "generative artificial intelligence" as a system trained on data that interacts through text, audio or visual communication and produces non-scripted outputs similar to human output with limited or no human oversight. The statute sets two different disclosure standards. First, for any act the Division of Consumer Protection administers, a person who uses generative AI to interact with someone must disclose, "if asked or prompted," that the other party is talking to generative AI rather than a human; using AI is also no defence to an underlying consumer-protection violation. Second, and more strictly, a person providing services in a state-regulated occupation must "prominently disclose" when a client is interacting with generative AI in that service, delivered verbally at the start of an oral exchange or through electronic messaging before a written one. The statute does not let AI substitute for a regulated occupation's own licensing requirements.

Character versus company

The obligation attaches to the business or professional deploying the system, not to the interactive persona's own scripted or generated dialogue. A virtual assistant with a name, voice and personality has no standing under this statute; the company operating it does, and it is the company's disclosure practice, not the character's in-world claims about itself, that the Division of Consumer Protection can act against. A persona that never volunteers its artificial nature unless a Utah user directly asks would meet the general standard, but would fail the stricter one if deployed inside a licensed profession.

What to watch

• Does a given virtual-persona deployment sit inside a state-regulated occupation, triggering the prominent-disclosure standard rather than the ask-only one?
• Has the Division of Consumer Protection brought or settled any enforcement action citing this section?
• Have other states adopted a similarly two-tiered disclosure structure since May 2024?

The statute's real significance for this site is the split it draws between a low bar for general AI interactions and a much higher one once a regulated profession is involved, a distinction worth checking before assuming any single disclosure standard applies.