New Here? Start Simple
A group of Utah lawmakers wrote a rule for companies that make AI chatbots and AI tools.
The rule said: if your AI is big enough and powerful enough, you have to tell the state what could go wrong with it, keep kids safer when they talk to your chatbot, and let your own employees speak up without getting fired if they see something dangerous. If a company broke the rule, the state could fine them a lot of money.
Almost everyone who voted on it — Republicans and Democrats — said yes. Then someone from the White House sent a letter asking Utah to stop. Utah's leaders never voted no on the bill in public. They just let it quietly die on the very last night, without ever explaining why.
The rest of this page shows exactly how that happened, and exactly what the bill would have protected.
Quick Glossary — Tap a Term
What It Was Supposed to Protect — and How
| What it protected | How the bill did it |
|---|---|
| Kids using AI chatbots | Any company running a chatbot kids could talk to had to have a written "child protection plan" in place. |
| The public, from hidden AI risks | The biggest AI companies (over $500 million a year in revenue) had to publish a public safety plan explaining the worst things their AI could realistically do. |
| The state's ability to catch problems early | Companies had to report safety incidents to a new state office — the Office of AI Policy — so problems didn't just stay secret inside the company. |
| Employees who see something dangerous | If a worker got fired or punished for reporting a safety problem, they could personally sue the company in court. |
| Actual accountability, not just a rule on paper | Real fines: up to $1,000,000 for a first violation, $3,000,000 for every violation after that, collected by the Attorney General. |
What the Bill Itself Would Have Done
| Part 1 — Transparency & Child Protection | |
|---|---|
| Who it applied to | "Large frontier developers" — AI companies with ≥$500M annual revenue building models trained on ~10^26 operations |
| Public safety plans | Required disclosure of catastrophic-risk assessments to the public |
| Incident reports | Filed with a new state Office of AI Policy |
| Child protection plan | Required for any developer operating a "covered chatbot" |
| Part 2 — Whistleblower Protection | |
| Employee right of action | Employees could personally sue for retaliation after reporting safety violations |
| Enforcement | |
| Who enforced it | The Attorney General |
| Penalties | Up to $1M first violation, $3M each violation after — deposited into a new AI Transparency Enforcement Restricted Account |
| The gap | Some incident reports/risk assessments could be withheld under existing GRAMA protected-record categories; safety-plan redaction also allowed for trade-secret, cybersecurity, or national-security reasons — a real weakness, not a blanket exemption |
A Bipartisan Committee, Then Silence
- Jan 27, 2026House Economic Development and Workforce Services Committee hears the bill. Witnesses in favor include Utah Parents United, Child First Policy Center, and Digital Childhood Institute; in opposition, TechNet and the Abundance Institute. The committee substitutes and passes it favorably, 8-0-2 — two members absent, zero votes against.
- Feb 5, 2026Reaches 3rd Reading. Majority Leader Casey Snider moves to "circle" the bill — pulling it from an immediate floor vote — after Fiefia discusses a teenager's death on the floor.
- Feb 12, 2026The White House Office of Intergovernmental Affairs sends a letter to Sen. Kirk Cullimore Jr. opposing the bill. White House officials separately hold direct conversations with Fiefia, telling him no changes would satisfy them.
- 26 daysThe bill sits with no floor action while the pressure campaign runs.
- Mar 3, 2026Sent from the 3rd Reading Calendar back to House Rules — three days before the session ends. The same day, 50+ Republican state legislators from 22 states send Trump a joint letter opposing the pressure campaign on federalism grounds.
- Mar 6, 2026, 11:59 PMEnacting clause struck by the Clerk of the House on the session's final night. No floor vote was ever recorded. It is the same kill mechanism already documented in this project's H.B. 468 case file — used here for a second time in as many sessions.
What the White House Actually Said the Problem Was
This is the entire stated justification. There isn't more of it.
Utah Public Radio reported plainly: the one-page letter "did not provide a legal justification for its opposition." No specific provision was named. No section of the bill was identified as the problem. "Unfixable" was never explained.
After Child First Policy Center put up billboards in Salt Lake City calling out White House AI adviser David Sacks by name, a spokesperson offered a second line of defense to Axios: the administration "fully supports child safety and has never told a State that it cannot enact child safety protections." That statement doesn't address the bulk of what H.B. 286 actually did — the frontier-model safety-plan and catastrophic-risk disclosure requirements that applied regardless of child safety.
The broader policy argument, offered separately by Sacks and echoed by Cullimore, Senate President Stuart Adams, and floor sponsor Sen. McKell, was about a "patchwork" of state AI laws: the fear that 50 different state rules would create compliance chaos for AI companies and slow the U.S. down in an AI race with China. Cullimore drew his own distinction publicly — he argued it's acceptable to regulate how AI is *used* by consumer-facing products, but not to regulate "the underlying technology" itself, which is what H.B. 286's frontier-model provisions did. That is a policy position, not a legal one, and Congress itself has spent years discussing federal AI legislation without passing any.
What "The Administration's AI Agenda" Actually Is
The White House letter cited an "Administration's AI Agenda" as the reason H.B. 286 was unfixable. Here's what that agenda actually consists of, in order. Only one item on this list is an actual law passed by Congress — everything else is executive action, which a future president could reverse with a signature.
- Jan. 23, 2025Executive Order 14179, "Removing Barriers to American Leadership in AI" — revokes the prior administration's AI executive order; sets deregulation as the starting posture.
- May 19, 2025TAKE IT DOWN Act (S.146) — signed into law. The only AI-specific federal statute actually enacted under this administration. Criminalizes knowing publication of nonconsensual AI-generated intimate imagery and requires platform takedowns within 48 hours of a valid request. Does not touch chatbot safety, frontier-model risk disclosure, or whistleblower protections — the subjects H.B. 286 covered.
- June 2025A second executive order amending/superseding EO 14179.
- July 2025"America's AI Action Plan" released — a policy document, not legislation, setting out the administration's deregulatory approach to AI development.
- July 23, 2025Executive Order 14319, "Preventing Woke AI in the Federal Government."
- Nov. 2025"Genesis Mission" launched — a Department of Energy AI-infrastructure initiative.
- Dec. 11, 2025Executive Order 14365, "Ensuring a National Policy Framework for Artificial Intelligence" — the order behind the H.B. 286 pressure campaign. Directs federal agencies to identify and challenge state AI laws seen as creating "an unworkable patchwork," and to consider conditioning discretionary federal grant funding on states that don't drop conflicting laws. Carves out state child-safety protections from preemption, but doesn't define how broadly that carve-out applies.
- June 2, 2026A second frontier-AI-security executive order, "Promoting Advanced Artificial Intelligence Innovation and Security" — signing delayed from May 21 amid industry pushback.
Six weeks after killing Utah's bill, the White House released its own national AI policy framework document on March 20, 2026 — asking Congress to eventually pass child-safety requirements for AI platforms, with no bill text and no timeline attached. Utah had a bill ready to sign that already did this.
A Question Worth Asking
If the goal is winning the AI race against China, how does requiring a chatbot company to have a child-safety plan slow that race down?
The "patchwork of state laws" argument makes sense for rules that touch how fast a company can train or deploy a frontier model — compute limits, development restrictions, things that could genuinely cost a company time or money in a global race. H.B. 286's child protection plan requirement wasn't that. It didn't restrict what a company could build. It required a plan for keeping kids safer when they talked to a chatbot that was already built and already on the market.
No one in this record — not the White House's letter, not Sacks, not Cullimore — has explained specifically how a child-safety plan requirement makes American AI development slower or less competitive. If there's an answer, it hasn't been made public. Readers can decide for themselves whether "national competitiveness" was really the reason this particular provision had to go.
A Second Question Worth Asking
Utah lawmakers substitute and rewrite bills constantly — including this one, once, in committee. So why didn't they rewrite H.B. 286 again to save what the White House would accept, instead of letting the whole thing die untouched?
H.B. 286 already went through one substitute (S1) in committee on Jan. 27, before any federal pressure existed. Utah's Legislature routinely narrows, reshapes, or guts-and-replaces bills mid-session to get something passable out the other end — it's a normal, frequently used tool here. During the 26 days the bill sat circled after the White House letter arrived, no second substitute was ever filed. No amendment. No attempt to strip out whatever the administration found "unfixable" and pass the rest.
McKell told Deseret News that if they had to "walk away" from the bill, they would still focus on AI child safety — but no replacement bill, narrower version, or child-safety-only substitute has been documented as introduced before the session ended. The Legislature had the tool it uses on other bills all the time. It didn't use it here.
Why the White House got involved
David Sacks, the administration's AI and crypto czar, is named by multiple outlets as driving opposition to state AI transparency bills generally; public reporting does not establish that he personally ordered the Feb. 12 letter to Cullimore. EO 14365 (above) carves out state child-safety protections from preemption — and H.B. 286 did include a child protection plan requirement — but the bill's larger provisions, requiring catastrophic-risk transparency for frontier AI models generally, fall outside that carve-out entirely. No outlet has published which specific provision made the bill "unfixable" in the administration's view.
Where Was the Federalism Commission?
Utah has an entire government body statutorily built for exactly this moment. The Federalism Commission — chaired by Sen. Keven Stratton and Rep. Ken Ivory — exists, per state law, to develop training on "state sovereignty" and "limits on federal government power." No discussion of H.B. 286 appears anywhere in the Commission's publicly available agendas or minutes.
What Schultz Says About Federalism
"The NFI is about pushing back against federal overreach, re-educating states on their rights, and getting back to the principles our founders intended."X, Sept. 26, 2025
"Every state deserves the authority to serve its own people."Facebook, Aug. 13
"Sign us up" — pitching Utah as a "national experiment" to "discontinue the federal oversight" on Medicaid, education, and public lands.ABC4, April 9, 2025
What He Said When It Actually Happened
Lawmakers would "continue studying the legislation" because he "understands the Trump administration's concerns."Utah Public Radio, on the H.B. 286 letter
No public statement invoking state sovereignty, federal overreach, or the Constitution in response to the White House killing an actual Utah law.
No record of Schultz raising H.B. 286 at any Federalism Commission meeting, summit, or public forum.
Federalism, in other words, gets invoked when the state is asking Washington for more control — not when Washington reaches in and overrides a law the state's own committee passed without a single no vote.
Access, Not Distance
Four months after H.B. 286 died, Schultz stood in the Oval Office on July 13, 2026, as Trump signed proclamations cutting Bears Ears by roughly 91% and Grand Staircase-Escalante by roughly 90% — an unrelated fight, but the same posture: present, supportive, photographed. The Salt Lake Tribune's own editorial board wrote afterward that Utah's political class should "be quiet and stay out of the photos."
Schultz has also personally hosted Donald Trump Jr. and Kimberly Guilfoyle for a weekend hunting trip at his own Croydon ranch, alongside Attorney General Sean Reyes and other Utah Republicans — access that predates H.B. 286's death but illustrates the same relationship: close, not adversarial, regardless of what the administration does to state law.
The Money Behind the Two Sponsors
The bill's two chief sponsors carried it into very different financial rooms. The records available don't establish that campaign money influenced either sponsor's actual position on H.B. 286 — they do establish that the two lawmakers carrying the same bill entered the debate from markedly different financial constituencies.
| Rep. Doug Fiefia (House sponsor) | Sen. Michael K. McKell (Senate sponsor) | |
|---|---|---|
| Primary funding base | Recurring individual donors | PACs, trade associations, corporate committees |
| Notable vendor spending | — | $16,500 to Irish Elk LLC; $20,000 to In The Field LLC |
| Public tone on the pressure campaign | "I appreciate the White House's engagement... and look forward to continuing the dialogue." | Called it "refreshing" for the White House to engage |
Graded against the standing rubric
This isn't a personal opinion of the bill. The grade below asks two questions, applied the same way to every bill in this series — not whether any one person likes the outcome: 1. Power — does it add or remove a check on power that the public, collectively, would otherwise have no say over? 2. Transparency — does it add or remove what the public can actually see about the process, regardless of who ends up holding the final decision? Design and outcome are graded separately because a bill can be structurally sound and still fail to become law — collapsing the two into one score would hide which one actually happened.
As Designed
A genuine transparency mandate with real enforcement teeth, undercut only by its own carve-outs.
As It Happened
A transparency bill killed in the dark, by a body that stayed silent when it mattered most.
Sources
Show sources ▸Hide sources ▾
Sources
Show sources ▸Hide sources ▾- H.B. 286 S1, official bill text, sponsor list, and status history — le.utah.gov
- House Economic Development and Workforce Services Committee, official minutes, Jan. 27, 2026
- Axios, "Scoop: White House pressures Utah lawmaker to kill AI transparency bill," Feb. 15, 2026, and "Utah billboards call out David Sacks over AI bill," Feb. 26, 2026
- Deseret News, "Utah lawmakers respond to Trump White House memo on AI bill," Feb. 17, 2026, and related coverage through March 2026
- Utah Public Radio, "Utah AI regulation bill stalled after opposition from the White House," March 18, 2026
- Model Republic, "Utah's AI bill is everything David Sacks asked for. He still wants it dead."
- PYMNTS.com, "Utah Becomes Site of Last Flashpoint Between States and the White House Over AI Regs"
- KSL.com, "Utah built a shield against AI harm; the White House knocked it down," March 16, 2026
- Governing magazine, "Utah's Blueprint for Governing Emerging Technology," Q3 2026 (Cullimore profile, S.B. 149, ALEC model legislation)
- Executive Order 14365, "Ensuring a National Policy Framework for Artificial Intelligence," Federal Register, Dec. 2025
- Utah Public Radio, Speaker Schultz statement on the White House letter
- ABC4/Inside Utah Politics (Lindsay Aerts), "State leaders tell Trump cabinet members that Utah should run its own federal programs," April 9, 2025
- Speaker Mike Schultz, X (@mschultz_12) and Facebook, National Federalism Initiative Summit posts
- KUER, Salt Lake Tribune, Utah News Dispatch — Bears Ears/Grand Staircase-Escalante Oval Office signing coverage, July 13, 2026
- KSL.com, Donald Trump Jr. Utah visit and Croydon ranch hunting trip coverage
- Utah Federalism Commission roster and statutory duties, Utah Code § 53H-4-703; National Federalism Initiative program materials, UVU Center for Constitutional Studies
- Campaign finance: Utah State Division of Elections disclosures, Fiefia and McKell committees