P07 / AI governance / Independent work sample
Shadow AI governance.
Unapproved AI use can expose sensitive business information.
The decision
behind the work.
Discover AI usage, classify destinations and prompt data, enforce egress policy, and manage justified exceptions.
Methodology and scope are defined in the project manifest. Listed capabilities describe the module design; source files show the implemented subset.
Assessment approach
- AI destination discovery
- Approved-tool registry
- Prompt data classification
- Allow-block-review decision
- MCP audit correlation
- Privacy-preserving event storage
Evidence
to request.
Start with source records, owner confirmation, scope and observation dates. A completed template alone does not establish operating effectiveness.
Defined output contract
- ai-destination-inventory.json
- egress-decisions.json
- exception-queue.json
- trend-series.json
The manifest defines these expected artifacts; confirm their existence and completion in source before relying on an output.
Read the full manifest ↗Review & decision boundaries
- Do not retain raw prompts by default
- Use data minimization and access controls
- Employee monitoring requires privacy and labor-law review
Remediation sequence
Record each finding with its evidence reference, risk rationale, accountable owner, target date and closure test. Escalate missing evidence rather than treating it as a pass.
Present management with the supported conclusion, remaining uncertainty and a specific decision request.