AI-103 Certification Practice Question #8

Question

You have a Microsoft Foundry project that contains a customer support agent built by using the Foundry Agent Service.
The agent uploads user-provided screenshots to Azure Storage through a ticketing tool and receives a blob URL for additional reasoning.
You need to use image moderation during agent runs and prevent harmful content from being returned during runs. Azure AI Content Safety must access the images by using the blob URL. The solution must follow the principle of least privilege.
What should you configure for Content Safety? To answer, select the appropriate options in the answer area.
NOTE: Each correct selection is worth one point.

Accepted Answer

Two selections are required. For storage access, Content Safety must read the screenshot via the blob URL; the documented mechanism is a managed identity granted a Storage Blob *Data* role — and MS Learn states explicitly that "Only Storage Blob Data Contributor or Storage Blob Data Owner are valid roles to proceed," so the Reader role does not work even though it appears more least-privilege. That eliminates access keys (no managed identity), the Storage Queue role (wrong data plane), and Storage Blob Data Reader (not accepted by Content Safety), leaving the system-assigned managed identity with Storage Blob Data Contributor. For guardrails, the requirement is to *prevent harmful content from being returned* during runs, which requires the Block action (not Annotate, which only flags), and it must cover the surfaces through which the moderated image content travels — user input, model output, the tool response carrying the blob URL/image, and the tool call. The option that selects all of those and sets Action to Block satisfies this; the narrower "Tool call only / Block" leaves the output and tool-response paths unguarded. Confidence is held at 78% because, while the storage-role selection is strongly documented, there is no community vote or Suggested Answer to confirm the precise guardrail surface set, and the broad "Block everything" guardrail choice is somewhat in tension with the stated least-privilege principle.

More AI-103 practice questions