Understanding Utah’s AI Sandbox

AI That Works for You, Safely

New AI tools are reaching Utah’s clinics, banks and classrooms faster than the rules written for them. Utah’s AI sandbox is where a company can test one of those tools under written conditions and state oversight, so we find out what it actually does before it reaches everyone.

What is a Regulatory Sandbox?

Its legal name is the Artificial Intelligence Learning Laboratory. Most people call it the sandbox, and that is a fair description: a small, supervised space to try something before it goes out into the world.

A company that wants to test an AI product here signs an agreement with this office and with the agency that already regulates its field. The agreement can adjust how one specific rule applies while the test runs. Everything else stays in force. In exchange, the company accepts limits on how far the test can go, safeguards it has to keep running, disclosures it owes the people using the product, and reporting to us.

It is also genuinely temporary. A first test period cannot run longer than 12 months, and a company can ask us to extend it twice, for 12 months each time. After that the agreement ends and the ordinary rules apply in full.

What every agreement has to spell out

  • How far the test can go, including how many people can use the product, who they are, and where.
  • The safeguards the company has to put in place.
  • Exactly which rule is being adjusted, and how.
  • What the company has to tell the people using the product.
  • What it has to report to us, so we can audit it.

Required by Utah Code § 13-72-401(4).

Our Commitments to You

The sandbox exists to answer questions about AI with evidence instead of guesswork, and to do it without using Utahns as the experiment.

  1. Observe and learn

    We do not take a company’s word for what its product does. We look at the data and the real-world performance, and we publish what we find so lawmakers and the public can see it too.

  2. Protect the public

    Safety is the entry price, not a bonus. Before a test starts we set limits on its scale, require safeguards and human oversight, and require the company to tell users what they are dealing with.

  3. Support innovation that earns trust

    Utah wants new tools that solve real problems here. A company can work out how an existing rule applies to its product with us and the responsible agency, in the open and on the record, instead of guessing.

How We Hold Participants Accountable

Getting into the sandbox is not the end of the review. It is the start of one.

Before an agreement is signed, a company has to satisfy us on five points: that it has the technical skill to run the product responsibly, that it has the money to meet its obligations during the test, that the likely benefits to consumers outweigh the risks of adjusting a rule, that it has a workable plan to monitor and reduce those risks, and that the size and length of the test are limited to match them. We consult the agency that regulates the field, and outside experts where we need them.

Once the test is running, we audit it. The reporting built into the agreement is what those audits run on. If a company breaks the law or breaks the agreement, we can remove it from the program immediately, and the usual civil and criminal penalties still apply. We can also end an agreement at any time. No company has a right to be in the sandbox.

What the sandbox is not

Being accepted into the sandbox is not an endorsement or approval of a company or its product by the State of Utah, and the state is not responsible for what a participant does. A participant stays subject to every legal requirement the agreement does not expressly adjust.

See Utah Code § 13-72-301, § 13-72-401, § 13-72-402 and § 13-72-403.

What would you like to do next?

  • View AI sandbox participants

    Every organization taking part, the pilots we have approved, and copies of the signed agreements.

    View

  • Apply to our AI sandbox

    What we ask for, how an application is reviewed, and what happens between first contact and a signed agreement.

    Apply

  • Browse our policy work

    Our current learning agenda, past research, and the findings we have published on specific AI policy questions.

    Browse

  • Engage with us

    Ask a question, share feedback on an AI policy issue, or find out how to join a focus group.

    Engage