What Utah’s Clinical AI Sandbox Reveals About Independent Oversight

Gorijavolu R, Jaiswal N, Kesselheim AS, Sakran JV, Hernandez-Boussard T, Celi LA - Nature Medicine

  • Regulation and Clinical Evidence

Summary:

Utah became the first state to authorize an AI sandbox for prescription renewals, but a transparency audit of the pilot found that neither the participating company nor the state would release the underlying safety data, performance metrics, or adverse event reports, with operational metrics classified as protected trade secrets. Because a regulatory framework cannot rely on a company as the sole source of safety evidence, the authors propose a Registry of Independent AI Auditors that would mandate external evaluation as a condition of sandbox participation while preserving commercial confidentiality.

VIEW SOURCE

Recent Work from PORTAL

Talk bubble graphics representing commentary and opinion.

How Congress Can Fix Expert Advisory Committees

Daval CJR, Kesselheim AS - STAT

  • Regulation and Clinical Evidence
Simple graphic of magnifying glass examining papers to signify reviewing research publications.

International Regulatory Status and Reimbursement Recommendations of Haematological Malignancy Drugs Withdrawn from the FDA Accelerated Approval Pathway

Mulsant SS, Kesselheim AS, Naci H - Journal of Cancer Policy

  • Price, Value, and Access
  • Regulation and Clinical Evidence
Talk bubble graphics representing commentary and opinion.

The Impact of Semantics in Medicines Regulation

Costa E, Kesselheim AS - Journal of Law, Medicine, & Ethics

  • Innovation Incentives and Competition
  • Regulation and Clinical Evidence