What you are evaluating
Scope this entry to Model Armor and its supported integrations. The overview describes single-request inspection rather than conversational memory, and identifies image screening as preview. Documents and images are not supported by every integration.
A useful evaluation context
Teams can evaluate managed content inspection while learning to verify enforcement and integration-specific gaps.
Documented capabilities
The vendor describes these capabilities in the linked sources. Availability depends on the product edition and supported environment.
- Templates configure filters for prompt attacks, sensitive information, harmful content and malicious URLs.
- Inspect-only mode supplies logging without stopping the interaction; blocking requires the corresponding enforcement configuration.
- Supported integrations include Google Cloud services and selected MCP request/response paths.
Where it fits in the work
- Select an integration and document its supported content types and size limits.
- Run paired cases in inspect-only mode with Cloud Logging enabled.
- Enable the intended blocking behavior and compare filter logs with the actual model and tool outcomes.
APPLY THE IDEA / ILLUSTRATIVE EXERCISE
Make the outcome observable.
Send a harmless query and a synthetic prompt attack through an instrumented test application.
Evidence to look for
Logs identify the template and filter result, and the application records whether the model was invoked in each enforcement mode.
Use synthetic data and an authorized test environment. Agree the scope and recovery steps before enabling enforcement.
Questions for your evaluation
- Does this integration inspect the documents or only extracted text?
- What happens to oversized or encoded content?
- How will the team distinguish a model refusal from a Model Armor block?