Architecture and data flow
Review ingestion, indexing, retrieval, model interaction, tools, connectors, trust boundaries, and deployment.
Exact scope, evidence, access, stakeholders, and acceptance criteria are agreed before delivery.
Review ingestion, indexing, retrieval, model interaction, tools, connectors, trust boundaries, and deployment.
Assess source permissions, user identity, filtering, tenancy, sensitive data, retention, and output handling.
Assess prompt injection, malicious content, data leakage, over-broad retrieval, insecure tools, and unsafe output use.
Review testing, logging, monitoring, incident response, change control, vendor dependencies, and human oversight.
Confirm the business driver, environment, stakeholders, boundaries, assumptions, and evidence needs.
Review the current state, evidence, architecture, risks, dependencies, and material gaps.
Develop the agreed controls, artifacts, recommendations, implementation, or operating routines.
Confirm completion, residual risk, ownership, reporting, and the next improvement priorities.