TL;DR Utah Gov. Spencer Cox signed Executive Order 2026-08 on October 6, 2026. It tells executive-branch agencies to avoid AI tools that do not follow the state's pro-human principles, keep a human in the loop for decisions affecting individual rights, document AI inputs and outputs, and train staff. It binds agencies, not private firms. Vendors selling AI to Utah should prepare answers now.
Utah has turned its "pro-human AI" slogan into an order that state agencies must follow. Executive Order 2026-08, titled "Establishing Utah's Pro-Human Approach to Artificial Intelligence in State Government," was signed by Gov. Spencer J. Cox on October 6, 2026 and took effect immediately.
For most small companies the order changes nothing today. For any company that sells software with AI features to a Utah agency, it adds a test: agencies "shall avoid AI technologies that do not adhere strictly to the foundational principles" of the state's Pro-Human AI Initiative. This page explains what the order says, section by section, and gives a vendor answer sheet you can fill in before a procurement officer asks.
Event date: October 6, 2026. First reports: October 6, 2026 (KSL, KUTV and ABC4 Utah, per Google News), followed by Utah News Dispatch, republished by Route Fifty on October 7. We read the signed order itself.
What the order says
The order is four pages. We read the signed PDF, which the governor's office published alongside its news release. Its structure:
| Section | What it does |
|---|---|
| 1. Purpose | Promote responsible AI use in government using the Pro-Human AI Initiative principles, invest in workforce training, and set guardrails for citizen-facing tools |
| 2. Definitions | "Agency" covers executive-branch departments, divisions and offices, including the State Tax Commission, the National Guard and the Board of Pardons and Parole |
| 3. Foundational principles | Agencies "shall avoid" AI tools that do not adhere strictly to the principles: human-guided and human-enhancing |
| 4. Department of Government Operations | Leads AI-driven process improvements, sets safeguards against "unauthorized or rogue AI agents," helps agencies with human-in-the-loop review, and develops the State-Endorsed Digital Identity (SEDI) Framework |
| 5. Agency duties | Staff complete AI literacy training, agencies look for cost-effective AI uses, and they document AI inputs, outputs and roles in decisions that affect the public |
| 6. Division of Technology Services | With the Division of Human Resource Management, builds the AI literacy curriculum, trains agency AI leads, and publishes uniform AI implementation standards |
| 7. Reporting | Progress report to the governor by July 1, 2027, then every July 1. Existing obligations under GRAMA, the Government Data Privacy Act and federal law are unchanged |
The definition of agency excludes a long list of bodies: higher education institutions, the Utah Board of Higher Education, the State Board of Education, independent entities as defined in Utah Code 63E-1-102, the Attorney General's Office, the State Auditor's Office, the State Treasurer's Office, and the legislative and judicial branches. If your customer is a university or a school board, this order does not reach them.
The two principles
The order defines its standard in two short paragraphs. "Human-Guided" AI should "protect human dignity, preserve human agency, and ensure individuals remain in control of the tools that shape their work and lives," and should be "understandable in how it functions, accountable in its impact, and adaptable to human needs." "Human-Enhancing" AI should "amplify human potential."
That is a values statement, not a test with pass marks. The practical requirements sit in sections 4 and 5.
The concrete duties
Three duties are specific enough to turn into vendor questions:
- Human review for rights-affecting decisions. The Department of Government Operations must help agencies deploy AI "while ensuring human-in-the-loop review for determinations affecting individual rights."
- Documentation. Each agency must maintain transparency "by documenting how AI systems, including their inputs, outputs, and roles in decision-making, are integrated into Agency processes that impact the public."
- Rogue agent safeguards. The department must "establish guardrails and technical safeguards to protect State information technology assets, systems, and data from cyberattacks, malicious actors, and unauthorized or rogue AI agents."
The order also tells the Division of Technology Services to prioritize "Utah-based entities" when it evaluates AI training platforms and content providers. That preference is limited to training, not to AI tools generally.
Who said what
In the governor's news release, Cox said: "The choice between innovation and safety in government is a false one." Marvin Dodge, Commissioner of the Utah Department of Government Operations and a Pro-Human AI Task Force member, said the state is "empowering our workforce to innovate safely." The initiative's co-chairs, Commerce Commissioner Margaret Woolley Busse and Governor's Office of Economic Development Commissioner Jefferson Moss, called the order "another vital action" for the effort.
The release says effective state government use of AI is one of seven pillars of the Pro-Human AI Initiative, which the governor established in December 2025. Utah News Dispatch, in coverage republished by Route Fifty, reported that the order requires agencies to use AI to "streamline internal processes and radically improve citizen-facing applications."
How it compares with other governor orders
Utah joins a busy month. Our governor AI executive orders tracker covers orders from California, Virginia, Illinois and Oregon issued between September 18 and 23. Utah's order differs in two ways. It is mostly about the state's own use of AI and staff training, not about overseeing frontier developers. And it names a concrete trust tool, the SEDI digital identity framework, which the order says should help prevent "identity fraud, impersonation, unauthorized AI agent activity, and fraudulent digital content."
Like the others, it does not set penalties for private companies. If you are tracking California's kill switch study, see our N-9-26 vendor clauses. For local rules that would reach private deployers, see the NYC Council AI bills tracker.
A separate Utah move came a day earlier. On October 5, the state's Office of Artificial Intelligence Policy announced healthcare agreements, including a master agreement with Intermountain Health and University of Utah Health and pilots with companies such as Nolla Health, according to Utah News Dispatch as republished by Route Fifty. That is the state's regulatory sandbox, which is a different program from this executive order.
What a vendor selling to Utah should prepare
If you sell a product with AI features to a Utah executive-branch agency, expect questions that map to sections 3 to 5. These are the questions we would prepare for. They are our reading of the order, not a form the state has published.
- Which decisions does your AI touch? Separate features that only draft or summarize from features that recommend or decide something about a person. The human-in-the-loop duty applies to determinations affecting individual rights.
- Where is the human checkpoint? Show where a staff member reviews, edits or overrides the output before it takes effect, and whether that step can be skipped.
- Can the agency document inputs and outputs? Agencies must document "inputs, outputs, and roles in decision-making." If your product cannot export logs or a plain description of what data goes in, the agency has to build that record itself.
- What stops an agent from acting outside its scope? If your product runs autonomous agents, describe permissions, rate limits and how an administrator stops one.
- What training do users need? Agencies are building AI literacy training. A short, accurate user guide helps your product fit into it.
Copy and adapt: vendor answer sheet
Fill this in once and attach it to proposals. Have counsel review it. Keep it factual; an inaccurate safety claim is worse than a blank.
Pro-Human AI Answer Sheet: [Product name], version [x.y], date [YYYY-MM-DD]
Reference: Utah Executive Order 2026-08 (signed October 6, 2026)
1. AI features in scope
- Feature: [name]. Purpose: [draft / summarize / recommend / decide].
- Does the output affect an individual's rights or benefits? [Yes / No]
2. Human review (EO 2026-08, section 4(b)(i))
- Where a person reviews the output before it takes effect: [describe].
- Can the review step be disabled? By whom? [describe].
- How a reviewer overrides or corrects the output: [describe].
3. Documentation of inputs, outputs and role (section 5(d))
- Data the AI receives: [fields / documents / none].
- Logs available to the agency: [what, retention period, export format].
- Plain-language description of the AI's role: [attached / link].
4. Safeguards against unauthorized or rogue agents (section 4(a))
- Can the AI take actions without a human click? [Yes / No].
- Permissions and limits: [describe].
- How an administrator suspends the AI feature: [describe, time to effect].
5. Training support (sections 5(a) and 6)
- User guide: [link]. Admin guide: [link].
- Known limitations users should be told about: [list].
6. Contact for AI questions
- Name, role, email: [fill in].
If you are not a Utah vendor
The same five questions work as an internal policy for any small team. Write down which AI uses touch decisions about people, where a person checks them, and what you log. Our AI acceptable use policy template has a place for each. If you sell into several states, keep one answer sheet and note state differences in the multi-state compliance plan. For general vendor questions, use the AI vendor due diligence checklist.
What we could not verify
We read the signed order and the governor's news release, which the governor's office published and Utah Policy reposted. The governor.utah.gov site blocked automated access, so we used the copy of the same PDF hosted by Utah Policy and the release text there. We could not open the KSL, KUTV or ABC4 reports, and rely on Route Fifty's republication of Utah News Dispatch as the independent report.
The order does not say how agencies will check whether a tool adheres to the pro-human principles, and no uniform implementation standards have been published yet. Section 6 tells the Division of Technology Services to coordinate them, with no date. We found no published procurement form or questionnaire tied to the order. We found no verified social post from the governor's office about the signing, so this page has no embed.
