The 2026 HIMSS conference highlighted a widening gap between rapid artificial intelligence evolution and the regulatory frameworks meant to govern them. Federal officials from the HHS and FDA expressed concerns that current oversight models—primarily designed for static software—cannot keep pace with agentic AI systems that possess the capacity for autonomous action and self-improvement. While the FDA has cleared over 1,300 AI-enabled medical devices, the shift toward models that evolve post-deployment necessitates a transition from one-time reviews to continuous, data-driven monitoring.

The regulatory landscape remains fragmented as state legislatures attempt to fill the void left by a deregulatory federal stance. This patchwork of local rules poses significant challenges for national healthcare providers, who must navigate varying definitions of liability and safety guardrails across state lines. Industry experts warn that without clear federal signals, the 'black box' nature of AI could lead to complex malpractice disputes when autonomous agents eventually make clinical errors.

To address these risks, agencies like ARPA-H are pioneering 'supervisory agents' designed to monitor clinical tools in real-time. This provides a layer of automated oversight where human resources are insufficient to scale. The approach aims to optimize human intervention, ensuring that as AI takes a larger role in cardiovascular care, safety and accountability remain structurally embedded within the technology's deployment.

At a big meeting in 2026 (HIMSS conference), experts said that AI is moving much faster than the rules we use to control it. Government groups that keep us safe (HHS and FDA) are worried. In the past, they made rules for computer programs that stay the same (static software). But now, new AI can learn and change by itself (agentic AI). Even though the government has already approved over 1,300 AI tools (AI-enabled medical devices), they realize they can't just check them once. They need to watch them every single day to see how they change.

Right now, the rules are like a messy puzzle. Because there isn't one big rule for the whole country, different states are making their own small rules. This is hard for big hospitals because they have to follow different laws depending on where they are. Experts are worried that AI can be like a 'hidden box' (black box) because we don't always know why it makes a choice. If an AI makes a mistake while helping a patient, it will be hard to figure out who is responsible (liability).

To help fix this, a special science group (ARPA-H) is creating 'AI Bosses' (supervisory agents). These are special programs designed to watch over the AI tools that doctors use for heart health (clinical cardiovascular care). Since there aren't enough humans to watch every AI all the time, these AI Bosses do it for us. This makes sure that as AI does more work, there is always a safety system (accountability) built right in to keep everyone safe.