Utah Mandates AI Receipts for Clinical Sandbox Participants

The memorandum of understanding requires participants to use Glacis technology to map AI operations to OVERT, an open standard for verifiable evidence. By generating these receipts, systems record whether safety protocols—such as deferring specific requests to human clinicians—are functioning as intended. Joe Braidwood, CEO of Glacis, noted that the requirement provides a concrete trail for every AI decision, moving beyond theoretical safeguards to documented operational proof.

The Office of Artificial Intelligence Policy will designate which vendors and systems must comply as it expands the state’s clinical AI program. This initiative brings together healthcare organizations and independent reviewers to assess pilot results. Crucially, the system allows for integrity checks without requiring the disclosure of raw prompts, model weights, or sensitive patient data. Instead, hashes and non-content identifiers link the evidence to specific clinical events. The Coalition for Health AI (CHAI) will also join the effort, serving as a convener for the independent auditors tasked with examining these signed records.