AI Insider Threat Program

Design an "insider threat program for AI" — monitoring, real-time detection, and response for AI agents modeled on the insider threat programs companies already run for employees. Covers layering your own agent-harness telemetry (e.g. Uber's open-source AI Detection and Response, ADR) with a model provider's classifier signals (Enterprise Frontier Safeguards) and an LLM gateway perimeter, intervening close to the moment a prompt is entered or a tool is called, and a continuous loop that discovers new model behaviors and builds safeguards against them. Also covers feeding long-horizon context (code, design docs, ticket history) into security reviews and threat models. Based on Uber CISO Philip Martin. Use when planning AI security architecture, agent monitoring, or making the case for monitoring frontier-model deployments.

uygnoey Updated 4 repo stars

File contents

uygnoey/skills-from-claude-blog commit b6a967c497

Frequently asked questions

npx get-skill add uygnoey/ai-insider-threat-program