Exam SC-500 Topic 1 Question 12 Discussion
Actual exam question for Microsoft's SC-500 exam
Question #: 12
Topic #: 1
Question #: 12
Topic #: 1
You have a Microsoft Foundry project that contains a model deployment named Deployment1.
Deployment1 contains an agent named Agent1 that uses an existing default guardrail configuration.
You discover that Agent1 generates tool calls that contain harmful language.
You need to ensure that Agent1 responses containing harmful content are prevented from running. The solution must prevent changes to the configuration of Deployment1.
What should you do?
Deployment1 contains an agent named Agent1 that uses an existing default guardrail configuration.
You discover that Agent1 generates tool calls that contain harmful language.
You need to ensure that Agent1 responses containing harmful content are prevented from running. The solution must prevent changes to the configuration of Deployment1.
What should you do?
Suggested Answer: B Vote an answer
To prevent the agent from executing tool calls that contain harmful language while strictly keeping the model deployment configuration unchanged, you must assign a custom guardrail directly to the agent.
The Core Problem
By default, an agent in Microsoft Foundry inherits the guardrail configuration of its underlying model deployment. However, model deployment guardrails typically only evaluate standard User Input and Output hooks. They do not evaluate the outbound payload of a tool execution.
Furthermore, modifying the model deployment's configuration is explicitly restricted by the requirements.
The Solution: Create and Assign an Agent Guardrail
Microsoft Foundry's guardrail framework includes a specialized four-point intervention architecture. Two of these points are exclusively available for agents: Tool call (Preview) and Tool response (Preview). Because an agent-assigned guardrail completely overrides and replaces the deployment-level guardrail for that agent's traffic, you can enforce tool-level scanning cleanly at the application boundary without altering the model deployment.
Reference:
https://learn.microsoft.com/en-us/azure/foundry/guardrails/how-to-create-guardrails
The Core Problem
By default, an agent in Microsoft Foundry inherits the guardrail configuration of its underlying model deployment. However, model deployment guardrails typically only evaluate standard User Input and Output hooks. They do not evaluate the outbound payload of a tool execution.
Furthermore, modifying the model deployment's configuration is explicitly restricted by the requirements.
The Solution: Create and Assign an Agent Guardrail
Microsoft Foundry's guardrail framework includes a specialized four-point intervention architecture. Two of these points are exclusively available for agents: Tool call (Preview) and Tool response (Preview). Because an agent-assigned guardrail completely overrides and replaces the deployment-level guardrail for that agent's traffic, you can enforce tool-level scanning cleanly at the application boundary without altering the model deployment.
Reference:
https://learn.microsoft.com/en-us/azure/foundry/guardrails/how-to-create-guardrails
by Aubrey at Jul 26, 2026, 11:40 AM
0
0
0
10
Comments
Upvoting a comment with a selected answer will also increase the vote count towards that answer by one. So if you see a comment that you already agree with, you can upvote it instead of posting a new comment.
Report Comment
Commenting
You can sign-up / login (it's free).