Framework for Quantifying System-Level AI Harms in Critical Infrastructure
August 26, 2026
This framework bridges the gap between model-level testing and systemic risk by linking structured hazard analysis with probabilistic system modeling. Using the UK's Real Time Gross Settlement system as a case study, it demonstrates how to derive AI-driven loss scenarios in complex sociotechnical environments.
HOW THIS AFFECTS YOU
●
builderThis helps you anticipate how individual model failures propagate through complex organizational workflows.
●
policyYou can use this to move from model-centric safety checks to evidence-based governance of critical infrastructure.