What you are evaluating
This profile emphasizes AI Model and Application Validation, with AI Runtime as an adjacent module. Confirm integration, supported models and entitlement; the management API labels MCP connection scanning beta, which is not established production coverage.
A useful evaluation context
Teams comparing predeployment AI testing with operational enforcement can trace the evidence between both stages.
Documented capabilities
The vendor describes these capabilities in the linked sources. Availability depends on the product edition and supported environment.
- Validation endpoints start assessments and retrieve results for registered models.
- Runtime management endpoints configure connections and policies and retrieve violation events.
- Separate runtime integration inspects application prompts and responses under configured guardrails.
Where it fits in the work
- Register an authorized test application and record the model and connection settings.
- Run a bounded validation and inspect a reproducible finding with the application owner.
- Apply the corresponding runtime policy in a lab and repeat both the triggering case and a harmless control.
APPLY THE IDEA / ILLUSTRATIVE EXERCISE
Make the outcome observable.
Evaluate a synthetic document assistant before and after enabling a policy for an injected instruction.
Evidence to look for
The record contains the original test case, validation result, configured policy and observable runtime outcome.
Use synthetic data and an authorized test environment. Agree the scope and recovery steps before enabling enforcement.
Questions for your evaluation
- Does validation exercise the whole application or only its model endpoint?
- Which traffic path actually enforces the runtime decision?
- Which agent or MCP capabilities are beta in the selected release?