AI-Native Insurance for Agentic AI: Pricing, Underwriting, and End-to-End Automation

Traditional cyber insurance is designed for information systems, focusing on data breaches and security failures. Agentic AI, however, introduces 'action risk'—the potential for AI systems to autonomously invoke tools, modify external environments, and execute workflows. This shift requires a new underwriting approach that treats AI not as a passive asset, but as an active operational participant. The paper proposes a formal mathematical framework to price and structure insurance for these systems, ensuring that insurance functions as both a risk-transfer mechanism and a regulatory tool to incentivize safer AI deployment.

The core of the proposed framework is the 'risk state,' a multidimensional vector that quantifies the exposure of an AI deployment. Instead of generic questionnaires, the model uses five observable variables: autonomy category (the agent's capability), operational authority (the frequency of autonomous execution), permission exposure (the scope of external systems the agent can touch), governance maturity (the quality of controls and auditability), and dependency concentration (reliance on shared infrastructure). By mapping these variables to event probabilities and loss severities, the framework allows insurers to optimize contract terms—such as premiums, deductibles, and coverage limits—while satisfying constraints for profitability and incentive compatibility.

The paper establishes that insurability is not universal but exists within a specific region of the risk-state space. A key finding is that the feasibility of fixed-term insurance contracts deteriorates as a system's exposure grows, highlighting the necessity of governance as a prerequisite for coverage. The framework also provides a pathway for operationalizing these contracts through continuous monitoring, automated claims validation, and human-in-the-loop escalation. A healthcare case study demonstrates how this model can be applied to real-world care-coordination agents, showing how specific operational choices directly influence insurance feasibility and pricing.