At a moment when regulators are under pressure to respond to rapid AI adoption without clear playbooks, a session led by Dr. Michael Cary, Professor at Duke University, focused attention on a more fundamental question: how governance, not technology, should shape regulatory action.
Speaking with Megan Wood, CEO and Registrar of the College and Association of Nurses of the Northwest Territories and Nunavut, Cary positioned AI as a present regulatory reality rather than a future concern. For boards and councils still orienting themselves, he cautioned against equating action with adoption. “I think that first and foremost, boards and councils need to give some consideration to moving to action,” he said, stressing that this did not mean “rushing to adopt a lot of tools,” but being intentional and “setting up good governance systems” first.
That framing mattered for regulators facing growing expectations to respond decisively to AI while maintaining public trust. Cary challenged the assumption that governance is primarily technical. Instead, he described it as a structured decision-making discipline rooted in values. Effective oversight begins with understanding where AI is already influencing decisions, from clinical tools to administrative processes, and asking what standards should apply.
A recurring theme came up, Cary warned against assuming that tools validated in large systems will translate cleanly across jurisdictions. Data drawn from one population may obscure risks or inequities in another. For regulators, particularly those overseeing smaller or distinct communities, local validation emerged as a core responsibility rather than an optional safeguard.
The conversation also surfaced concerns about fairness for registrants. Cary acknowledged the risk of biased historical data influencing automated flags or assessments, and emphasized guardrails such as diverse governance bodies, bias assessments across sub populations, and ongoing monitoring as tools evolve. Importantly, he reinforced that AI should support human judgment, not replace it.
For regulators in the room, the session clarified that AI governance is not an add on to existing mandates. It sits squarely within them.
