A model does not operate alone: it relies on identity services, stored context, retrieval tools, connected applications and monitoring. Mapping those links shows which component an attacker could misuse and where a separate control must interrupt the path.
Eighteen governed components around one model. Select any component for its definition, assurance risk, governing control, accountability and evidence. Trace the operating flow or the evidence flow one relationship at a time. This is the evidence and lifecycle portion of the AI Harness lab, presented on its own so it can be expanded and explored.