#cloud

Cloud / AI notes: testing the trust boundary of an internal assistant

A lightweight threat model for an assistant using internal data: what is trusted, and which actions need human approval.