Zero-Trust · Distributed PEPs → Central PDP
A zero-trust architecture for enterprise AI
A navigable reference framework: never trust, always verify, assume breach — applied to AI agents and applications.
Request flow (North-South) Control / policy flow (East-West) Telemetry / response flow
The three principles
Never trust, always verify
Every request is authenticated and authorized, regardless of origin.
Assume breach
Design expecting compromise: limit blast radius, segment by identity.
Least privilege & least agency
Grant only what is needed — and constrain what each agent tool can do.