DataPacts

Utah: UCPA and AI Training Data

Checked

At a glance

Utah regulates AI training data at two points: the Utah Consumer Privacy Act (UCPA, Utah Code Title 13, Chapter 61) treats exchanging personal data for monetary consideration as a sale, and the Artificial Intelligence Policy Act (Title 13, Chapter 72, with the 2025 disclosure rules now in chapter 13-77) regulates how businesses disclose generative AI to consumers without any rule on acquiring or licensing training data.

Independent reference. Every buyer is covered on the same terms, every figure carries its source and check date, and no buyer pays to be listed or placed higher.

Request an independent introduction

Tell us what data your company holds. We reply with the buyers that fit and what their terms leave out. Your details reach a buyer only after you agree to a named introduction.

IndependentYour details stay with us until you agreeNot selling is a valid outcome

The UCPA has applied since 2023-12-31, as of 2026-11-17, writes synthetic data into its definition of deidentified data and never mentions AI training. A data holder meets the UCPA and an AI developer the AI Policy Act as a deployer, a state layer on top of federal law within AI training data licensing.

Every rule below cites Utah Code or an enrolled bill read on 2026-11-17 (how we source, date and check every figure).

Two Utah Instruments on a Training Data Deal at a Glance

Two Utah instruments decide what a holder may license and what a developer must disclose, and the table gives each one's citation, start date and enforcement as of 2026-11-17.

Instrument Citation In force Who it binds What it requires in a data deal Enforcement Source
Utah Consumer Privacy Act (UCPA) Utah Code 13-61-101 et seq. (Laws of Utah 2022, ch. 462); 2026 amendments by ch. 95 and ch. 193 2023-12-31 (motor vehicle manufacturer provisions 2027-01-01) controllers and processors with revenue of USD 25,000,000 or more above the consumer-count thresholds licensing personal data for money is a sale; disclosure and opt-out; sensitive data after notice and opt-out; deidentified data includes synthetic data under three conditions attorney general only; 30-day cure; damages and up to USD 7,500 per violation; no private action le.utah.gov, Chapter 13-61
Artificial Intelligence Policy Act (S.B. 149, 2024) as amended in 2025 and 2026 Utah Code Title 13, Chapter 72; chapter 13-77 (Laws of Utah 2025, ch. 465, enrolled as S.B. 226 sections 13-75-101 to 106); 63I-2-213(2) (S.B. 332, 2025) 2024-05-01; disclosure rules from 2025-05-07; repeal date 2027-07-01 businesses and regulated-occupation providers that use generative AI with consumers; the Office of Artificial Intelligence Policy and learning laboratory participants no rule on acquiring or licensing training data; AI disclosure when a consumer asks; high-risk interactions disclosed Division of Consumer Protection; fines up to USD 2,500 per violation le.utah.gov, Chapters 13-72 and 13-77, 63I-2; enrolled S.B. 226 and S.B. 332

The sections follow a deal: sale, deidentified data, in-house use, then the AI Policy Act.

When Licensing Personal Data for AI Training Is a Sale Under the UCPA

Licensing personal data to an AI developer for money is a sale under the UCPA, which defines it as "the exchange of personal data for monetary consideration by a controller to a third party" (Utah Code 13-61-101). Three points follow.

  • The definition excludes a disclosure to a processor or an affiliate, a disclosure consistent with the consumer's reasonable expectations, a consumer-directed disclosure, and a transfer "as an asset that is part of a proposed or actual merger, an acquisition, or a bankruptcy in which the third party assumes control of all or part of the controller's assets".
  • A consumer may opt out of the sale (13-61-201(5)), and a controller that sells must disclose how (13-61-302(1)(b)).
  • A contract term that waives or limits a consumer right is void (13-61-302(6)).

The definition says "monetary" and reads no further, and no reading of non-cash deals was found as of 2026-11-17. The bankruptcy carve-out sits beside US bankruptcy data sales, where federal court conditions apply.

Which Businesses and Data the UCPA Covers

The UCPA applies to a controller or processor that conducts business in Utah or targets Utah consumers, has annual revenue of USD 25,000,000 or more, and either controls or processes personal data of 100,000 or more consumers in a calendar year or derives over 50 percent of gross revenue from selling personal data while handling the data of 25,000 or more consumers (13-61-102(1)), as of 2026-11-17. The chapter excludes governmental entities, higher education, nonprofit corporations, HIPAA covered entities and protected health information, Gramm-Leach-Bliley financial institutions and data, and Fair Credit Reporting Act activity (13-61-102(2)), so a smaller holder may sit outside it.

"Personal data" excludes deidentified data, aggregated data and publicly available information (13-61-101), and health data is covered under HIPAA and AI training data.

Staff and Business Contacts Fall Outside "Consumer"

The UCPA protects a "consumer", an individual who is a Utah resident acting in an individual or household context, and the term "does not include an individual acting in an employment or commercial context" (13-61-101).

Staff archives and B2B contact records therefore fall outside the consumer rights, though contracts and other laws may still limit a license. Notice and consent for staff sit on the page about employee notice and consent when licensing internal communications.

A Utah controller may not process sensitive data collected from a consumer without first presenting clear notice and an opportunity to opt out (13-61-302(3)(a)), a lighter rule than the opt-in consent of Virginia or Texas.

Sensitive data covers racial or ethnic origin, religious beliefs, sexual orientation, citizenship or immigration status, medical history or diagnosis, genetic or biometric data processed to identify a specific individual, and specific geolocation (13-61-101).

"Biometric data" excludes photographs, video and audio recordings and data generated from them, and a known child's data follows COPPA (13-61-302(3)(b)). The comparison across states sits on the page about biometric data in AI training compared across laws.

Synthetic and Deidentified Data in the Utah Definition

The UCPA writes synthetic data into its definition of deidentified data: "Deidentified data includes synthetic data" (13-61-101), and synthetic data is "data that has been generated by computer algorithms or statistical models and does not contain personal data". Deidentified data carries three conditions.

  • The data cannot reasonably be linked to an identifiable individual, and the controller takes reasonable measures against association.
  • The controller publicly commits to keep the data deidentified and not to reidentify it.
  • The controller contractually obligates every recipient to comply.

A controller using deidentified data must take reasonable steps to comply with its contractual obligations and promptly address a breach of them (13-61-303(3)), and nothing requires reidentification (13-61-303(1)(a)).

The recipient duty puts part of the standard in the license, and the methods behind the first condition are set out under PII removal and de-identification before licensing data.

What the UCPA Says About Using Data to Develop Technology In-house

The UCPA never mentions artificial intelligence or model training, and its limitations section lets a controller or processor process personal data to "conduct internal analytics or other research to develop, improve, or repair a controller's or processor's product, service, or technology" (13-61-304(1)(l)(i)).

Public or peer-reviewed research in the public interest is also permitted if disclosed in the privacy notice (13-61-304(1)(j)).

A controller using an exemption bears the burden of showing it applies (13-61-304(4)). The statute does not say whether training a model falls under (l)(i), and no attorney general guidance or court reading was found as of 2026-11-17, so this page concludes nothing.

What the Utah AI Policy Act Regulates and What It Leaves Out

The Utah Artificial Intelligence Policy Act (Title 13, Chapter 72, enacted in the 2024 General Session as S.B. 149 and in force from 2024-05-01) creates the Office of Artificial Intelligence Policy and a learning laboratory, and the chapter text read on 2026-11-17 contains no rule on acquiring or licensing training data.

The only occurrence of "trained on data" sits inside the generative AI definition of chapter 13-77, which describes the system a business deploys. Laws of Utah 2026, ch. 127 renumbered and amended the chapter.

Disclosure Duties in Chapter 13-77, Enacted in 2025

S.B. 226 (2025 General Session, effective 2025-05-07) repealed the original disclosure section 13-2-12 and enacted the disclosure rules, which the code now carries as chapter 13-77: a supplier using generative AI in a consumer transaction must disclose that it is not human when the individual clearly and unambiguously asks, and a person in a regulated occupation must disclose high-risk generative AI interactions up front (13-77-103).

A system that discloses at the outset and throughout has a safe harbor (13-77-104). The Division of Consumer Protection enforces, with an administrative fine or a court fine of up to USD 2,500 per violation (13-77-105), as of 2026-11-17. None of it regulates what a model was trained on, so it binds a developer as a deployer only.

The Office of AI Policy and the Repeal Date

The Office of Artificial Intelligence Policy sits in the Department of Commerce (13-72-201) and runs a learning laboratory whose regulatory mitigation agreements have a demonstration period of up to 12 months, and a participant may request a 12-month extension, of which the office may grant up to two (13-72-401 to 13-72-403).

S.B. 332 (2025) extended the repeal date, and 63I-2-213(2) as read on 2026-11-17 states "Title 13, Chapter 72, Artificial Intelligence Policy Act, is repealed July 1, 2027".

How the Utah Rules Fit One Data License

A Utah data license passes four checks in order: the UCPA gate, personal versus deidentified or synthetic data, sensitive data, and what the buyer will deploy.

  1. Gate: revenue and consumer-count thresholds, as of 2026-11-17 (13-61-102(1)).
  2. Personal or deidentified: a sale needs disclosure and an opt-out; deidentified data needs the three conditions and a flow-down clause (13-61-302(1)(b), 13-61-101).
  3. Sensitive data: clear notice and an opportunity to opt out (13-61-302(3)).
  4. Developer side: the AI Policy Act binds a deployer's disclosures, not the dataset (13-77-103).

The license should record the deidentification commitment and the flow-down, and the clause detail sits in the page on AI data licensing agreements.

Benefits, Risks, Downsides and Alternatives for a Seller Under Utah Law, Including Not Selling

Utah law lets a company license data for AI training, but personal, sensitive and deidentified data each carry a condition, and licensing only synthetic or deidentified data or not licensing at all remain valid choices.

Benefits

  • A holder under the revenue gate may sit outside the Act (13-61-102(1)).
  • Staff and B2B data fall outside "consumer" (13-61-101).
  • Sensitive data needs notice and opt-out, not consent (13-61-302(3)).
  • A 30-day cure period precedes any action, and no private action exists (13-61-402, 13-61-305).

Risks

  • A sale needs disclosure and an opt-out (13-61-302(1)(b)).
  • Deidentification fails if one of the three conditions is missing (13-61-101).
  • No reading exists on whether in-house training falls under 13-61-304(1)(l)(i).

Downsides

  • Flow-down duties reach every recipient of deidentified data.
  • The controller carries the exemption burden (13-61-304(4)).
  • After a failed cure the attorney general may recover damages and up to USD 7,500 per violation (13-61-402), as of 2026-11-17.

Alternatives

  • License only deidentified or synthetic data under the three conditions.
  • Exclude sensitive data from the set.
  • Wait for guidance, with no result implied.
  • Not selling or licensing at all, which the alternatives to selling data to AI labs page lays out.

Beyond Utah: Other States, Federal Rules and the Seller Process

Utah is one state layer on top of federal law, and the seller pages turn these rules into steps. The page on United States AI training data law compares Utah with eleven other states.

How This Page Is Sourced and Dated

Every rule here cites a Utah Code section or an enrolled bill, with the date it was read. The method for sourcing and dating each figure is public. This page is general information about Utah law, not legal advice for any specific deal.

Frequently Asked Questions

Yes, with conditions. Licensing personal data for money is a UCPA sale with notice and an opt-out, and the Act applies only above a revenue gate of USD 25,000,000, as of 2026-11-17.

Does the Utah AI Policy Act regulate AI training data?

The chapter text read contains no rule on acquiring or licensing training data. Its disclosure duties concern generative AI used with consumers.

Does the UCPA protect employee data?

Employees acting in an employment context are not consumers under the UCPA. Contracts and other laws may still limit a license.

Can Utahns sue under the UCPA?

No: a UCPA violation gives no private right of action. The attorney general enforces after a 30-day cure period.