Polsia scores agentic AI risk

- Polsia proposed a quantitative risk score for agentic AI that ranks autonomy, data sensitivity, financial exposure, and audit recourse for governance controls. - Higher scores trigger human-in-the-loop requirements, escrowed audit logs, and stricter vendor attestations and annual red-team certifications by auditors. - The metric lets boards classify agents and set clear escalation thresholds for production deployments. (x.com)