NAIC Spring Meeting Flags Agentic AI as New Governance Risk

The NAIC’s Spring 2026 National Meeting in San Diego dedicated formal attention to agentic AI, defined as systems that pursue goals autonomously across multiple steps, as a distinct category of risk for insurers. According to summaries of the meeting, regulators identified accountability gaps, error propagation across chained agents, and performance limitations driven by data or technical constraints as the primary concerns.

The discussion is significant because the NAIC AI Systems Evaluation Tool already targets high-risk AI systems, and agentic claims or underwriting tools are likely to fall into that category. Carriers that treat agentic AI as just another automation layer risk missing the governance implications. When an AI system can initiate actions, retry failed steps, or coordinate across other systems without a human in each loop, the points of accountability shift from the model to the workflow design.

Regulators proposed a risk taxonomy with varying levels to help prioritize high-risk use cases, which suggests that future examinations may classify agentic tools separately from simpler predictive models. The guidance emerging from the meeting calls for monitoring agent use, establishing clear accountability, redesigning governance frameworks for agentic systems, and maintaining human-in-the-loop escalation for high-risk cases.

For carriers, the immediate action is to inventory where agentic tools are already in use or in pilot. Claims triage, customer service bots, and underwriting recommendation engines are the most common early deployments. Each should be reviewed against the same governance standards as high-risk models: documentation of decision logic, override mechanisms, monitoring logs, and escalation paths. The NAIC has not yet issued a dedicated agentic AI framework, but the Spring 2026 meeting made clear that one is coming.

For more on how regulation is targeting high-risk AI systems, see our guide to the NAIC AI Evaluation Tool.