Utah split chatbot disclosure into on request and by default

The conversation

Utah's Senate Bill 149, the Artificial Intelligence Policy Act, was signed by the governor on 13 March 2024, as the legislature's own bill tracking page records, and, by its own final section, took effect on 1 May 2024. The enrolled bill text creates a two-tier disclosure duty rather than a single blanket rule, tying the level of obligation to whether the interaction touches a state-regulated occupation.

What the documents show

The statute's own operative text states that a person who uses generative AI to interact with someone in connection with an act the state's Division of Consumer Protection administers must ‘clearly and conspicuously disclose’ that the person is interacting with generative AI, but only ‘if asked or prompted.’ A separate provision reaches further: a person providing the services of a regulated occupation, meaning one licensed under the state's professional-licensing title, must ‘prominently disclose’ the use of generative AI by default, without waiting to be asked, and that disclosure must be given verbally at the start of an oral exchange or through electronic messaging before a written one begins. The statute is explicit that meeting this disclosure duty does not excuse a person from otherwise meeting the substantive requirements of the regulated occupation itself.

The system boundary

Enforcement runs through the state's existing consumer-protection machinery rather than a new AI-specific body: the bill's text assigns administration to the Division of Consumer Protection and sets an administrative fine of up to 2,500 dollars per violation, with a court able to add injunctive relief, disgorgement and a further fine of up to 2,500 dollars per violation, and a civil penalty of up to 5,000 dollars for violating an order already issued. That structure marks the human escalation path the statute contemplates: a division investigation and, if needed, a court action, rather than automatic penalties triggered by non-disclosure alone.

Where it fails

Because the general disclosure duty applies only ‘if asked,’ a person interacting with a generative AI system outside a regulated occupation, and who never asks, may not receive disclosure at all under this statute; the by-default duty is confined to the licensed-occupation context the statute names. A builder operating outside Utah's regulated-occupation list should not assume this statute requires proactive disclosure for their product.

• Does this deployment fall within a Title 58 regulated occupation, which triggers Utah's by-default disclosure duty, or outside it, where disclosure is on-request only?
• If by-default disclosure applies, is it delivered verbally at the start of a spoken exchange or through electronic messaging before a written one, as the statute specifies?
• Is there a documented path for the Division of Consumer Protection's inquiries, given that enforcement runs through that division rather than automatically?

This is a plain reading of the enacted text; the statute does not itself define every term with examples, so edge cases outside a licensed profession may require separate legal interpretation.