Runtime decisions with a review trail
Place an explicit governance decision between an AI request and its effect, with policy versions and outcomes preserved for review.
01
A policy decision at the point of action
Each governed request is designed to follow the same decision stages.
- Identify the tenant, system, component, and active policy version.
- Evaluate the request and context against applicable controls.
- Return an allow, warn, redact, rate-limit, require-approval, or deny effect.
- Record a bounded decision result for later review.
Policies can express an immediate effect or pause an action for human confirmation. The application receives the governed outcome through the gateway boundary.
02
Human control remains explicit
A require-approval decision is designed to create a bounded approval request rather than silently weakening or bypassing the policy result.