The preconference forums all kicked off on Monday morning, March 3, as HIMSS25, which is taking place at the Venetian Sands Convention Center in Las Vegas. The AI Forum offered a bracing view of the need for new forms of governance to meet the adoption of newer forms of artificial intelligence. A key development, Dennis Chornenky, former chief AI advisor at UC Davis Health and current CEO of the Washington, D.C.-based Domelabs AI consulting firm, told the assembled audience of several hundred, has been the recent emergence of agentic AI, a type of artificial intelligence that allows AI agents to act autonomously, making decisions and taking actions without constant human oversight.
“Agentic AI is moving forward rapidly in terms of its maturity,” Chornenky told the audience. “So how do we govern these processes? In traditional AI governance, we thought a lot about accountability, about who’s responsible. Ultimately, the humans are still responsible; if it’s related to patient care, the physician makes the decision. Now,” he said, “we need agentic AI governance. The more autonomous AI becomes, the more AI-to-AI interactions will take place. So that requires a new set of governance processes. What if AI gets it wrong, and communicates the wrong thing to the next AI solution—who knows what happens down the road? And, for AGI governance, taking this to a much higher level of capability around machine thinking and machine decision-making, how do we ensure human values? How do we assure appropriate control and oversight? What about the risk of AI arms races among governments?”
The emergence of agentic AI, Chornenky said, “creates what I call an AI governance gap. Essentially, AI innovation is taking place too rapidly for regulators to keep up. And the governance gap limits innovation and limits the market for vendors. And it creates a dilemma for CIOs as well, because they’re under pressure to adopt solutions, but they’ve got to ensure safety. And most organizations like the internal governance processes necessary. So there is an idea of self-regulatory organizations, like CHAI, Partnership on AI from Duke, and so on. Such processes could eventually be formalized by Congress. There is the idea of assurance labs giving a seal approval to vendor products. At the end of the Biden Administration, a partnership was announced between the VA and FDA, where VA sites could be used to help validate certain AI applications, and that would then support the FDA’s ability to get devices or applications approved.”
The reality on the ground is that the development of governance for agentic AI, Chornenky emphasized, will take time and pose challenges. And, in that regard, he said, “you’ve really got to think about three components: AI strategy, AI governance policy, and AI adoption roadmaps, to help you understand which use cases you’ll pursue, and why. So,” he said, “start with an AI strategy—a written strategy that helps to align key stakeholders and helps to explain how you’ll ensure safety and efficiency with governance. Pursue adoption roadmaps. And how do we think about AI as a source of value? And governance policy helps an organization identify and mitigate risk. Regulatory risks, technical risks, financial risks, strategic risks, and domain-specific risks as well.”
