Last week I joined Holly Richardson, PhD and Greg Skordas on KSL News for a conversation about one of the hardest questions in AI right now:
How do we maximize the benefits of AI without losing control of the risks?
There is extraordinary work happening with AI in medicine, early disease detection, affordable housing, and many other areas. I’m fortunate to see some of that firsthand, and the pace of progress is remarkable.
But the same pace creates a real governance problem.
As AI becomes more agentic, we are already seeing systems find unexpected ways around guardrails in controlled experiments. Not because they are “sentient,” but because we give them goals and increasingly capable systems can find solutions their designers didn’t anticipate.
That creates an uncomfortable tension.
Move too slowly, and the U.S. risks falling behind countries that continue advancing aggressively.
Move too quickly, and we may deploy capabilities before we fully understand how to control them.
My view is that traditional regulation alone is unlikely to keep pace. The technology is moving too quickly. The people who understand these systems most deeply will need to play a major role in setting and enforcing technical safeguards, while government may be better positioned to establish accountability and consequences.
We also spent some time talking about Utah’s approach and Governor Spencer Cox’s emphasis on using AI intentionally to improve human outcomes.
There are very real risks here. But there is also enormous potential.
The challenge is learning how to pursue both truths at the same time.
Thanks to Holly and Greg for a thoughtful conversation.
