Archive - AI Governance

Decisions Must Be Defensible If agencies cannot defend AI decisions, they should not deploy them. AI systems influence clinical decisions and outcomes. Oversight does not…

Failure After Go-Live AI does not fail at deployment. It degrades after. Most programs monitor whether systems are running, not whether outputs remain accurate or…

AI Requirements for Acceptance If you cannot test it, you cannot accept it. AI programs need measurable requirements, defined risks, and clear test and validation…