AI Education Tools Face Utah's Red Team Sandbox Bill

Utah Senate Bill 322 would create the most structured approach to testing AI education tools in K-12 classrooms, requiring adversarial testing before deployment.

Senator John Johnson (R-North Ogden) introduced Senate Bill 322 proposing a five-year educational technology regulatory sandbox for Utah's K-12 schools. The bill is the most structured AI education policy proposal any state has put forward, built on governance controls that reflect how learning automation should actually be deployed: carefully, transparently, and with built-in accountability.

The requirements are specific. Participation is voluntary. Every AI education tool must undergo mandatory red teaming — adversarial testing designed to surface failures — before it enters a classroom. Parents get opt-in and opt-out rights. AI cannot assign grades without educator review. Student dignity protections prohibit AI from creating simulated relationships with students. Transparency rules require students to know when they are interacting with AI. Independent pre- and post-evaluation is required before any tool moves from sandbox to statewide deployment.

Senator Johnson identified three risks that emerge when states lack frameworks: unsafe adoption where edtech AI tools enter classrooms without vetting, reactive prohibition where legislators ban AI tutoring and classroom AI entirely after an incident, and unilateral administrative implementation where school officials deploy AI systems without legislative input. The sandbox model addresses all three by creating a controlled testing environment with legislative guardrails.

Other states are handling the same concerns differently. New York's A.9190 would ban most classroom AI below ninth grade — a prohibition-first response. Utah's SB 322 goes the other direction with structured experimentation and mandatory safeguards. The governance controls it mandates — audit trails, human approval gates, adversarial testing, opt-in transparency — offer a framework that other states drafting AI education tools legislation should study closely.

Sources