From Agent Failure Paths to Quantified Residual Risk: A Compositional Framework for Resilient Agentic AI

Short summary

CPSAINT is a seven-layer integrity decomposition framework for agentic AI that pairs physical state, sensors, data, compute, actuators, environment, and time with FRIESA-K, a residual-risk functional mapping failure paths to quantified risk. It grounds resistance in a controlled absorbing Markov model rather than informal scoring. The framework is demonstrated on a warehouse robot and a governance-instrumented financial-services agent, showing the same layer grammar and variable semantics hold across domains.

• •Seven-layer integrity decomposition (CPSAINT) paired with FRIESA-K residual-risk functional for quantified risk
• •Resistance term K grounded in absorbing Markov model rather than informal scores
• •Validated on contrasting scenarios: hard real-time warehouse robot and financial-services governance agent

Generated with AI, which can make mistakes.