Pass Microsoft AI-103 Exam in First Attempt Guaranteed!
Get 100% Latest Exam Questions, Accurate & Verified Answers!
30 Days Free Updates, Instant Download!
Microsoft AI-103 Practice Test 38
Question 370
Production requirement: A support bot reads cloud deployment status, and administrators want only a separate role to restart production services. Which TWO decisions would undermine the required system boundary? Select TWO answers.
Explanation
Correct Answer
“Execute any generated payment or deletion call without a human approval checkpoint” would violate the requirement.
“Expose a general administrator tool to all customers because arguments are described in the prompt” would violate the requirement. The appropriate control instead is to define a narrow validated tool schema, restrict its identity permissions and require explicit authorization for irreversible or high-impact tool calls.
Incorrect Answers
“Inspect records that explain the affected process, then check the result described in this case: A support bot reads cloud deployment status” adds a relevant safeguard or verification step rather than weakening the solution. Confirm that authorized read requests succeed; unapproved writes, cross-tenant identifiers and malformed arguments are rejected with traceable policy reasons.
“Authorized read requests succeed; unapproved writes, cross-tenant identifiers and malformed arguments are rejected with traceable policy reasons” is the required outcome of a working control, not an action that weakens it. Confirm that authorized read requests succeed; unapproved writes, cross-tenant identifiers and malformed arguments are rejected with traceable policy reasons.
“Define a narrow validated tool schema, restrict its identity permissions and require explicit authorization for irreversible or high-impact tool calls” is the controlled implementation, not an unsafe workaround.
Question 371
A team is investigating this task: A procurement agent may draft purchase orders, but any submission above a threshold requires verifiable approval. Which design decision is justified by the failure and the required outcome?
Explanation
Correct Answer
A procurement agent may draft purchase orders, but any submission above a threshold requires verifiable approval. A model-proposed tool call is untrusted until its arguments, caller authority, resource boundary and side-effect class are checked by the application. Authorized read requests succeed; unapproved writes, cross-tenant identifiers and malformed arguments are rejected with traceable policy reasons.
Incorrect Answers
A team is investigating this task: A procurement agent may draft purchase orders, but any submission above a threshold requires verifiable approval. “Pass user-submitted arbitrary resource IDs directly into a privileged action” does not implement the control that the facts call for. A model-proposed tool call is untrusted until its arguments, caller authority, resource boundary and side-effect class are checked by the application. Instead, define a narrow validated tool schema, restrict its identity permissions and require explicit authorization for irreversible or high-impact tool calls.
A procurement agent may draft purchase orders, but any submission above a threshold requires verifiable approval. “Trust model JSON syntax alone as authorization” does not implement the control that the facts call for. A model-proposed tool call is untrusted until its arguments, caller authority, resource boundary and side-effect class are checked by the application. Instead, define a narrow validated tool schema, restrict its identity permissions and require explicit authorization for irreversible or high-impact tool calls.
“Apply the proposed setting to all connected workloads before testing the case-specific boundary: a procurement agent may draft purchase orders” does not implement the control that the facts call for. A model-proposed tool call is untrusted until its arguments, caller authority, resource boundary and side-effect class are checked by the application. Instead, define a narrow validated tool schema, restrict its identity permissions and require explicit authorization for irreversible or high-impact tool calls.
Question 372
Situation: A procurement agent may draft purchase orders, but any submission above a threshold requires verifiable approval. Which underlying technical principle explains the recommended implementation?
Explanation
Correct Answer
A procurement agent may draft purchase orders, but any submission above a threshold requires verifiable approval. A model-proposed tool call is untrusted until its arguments, caller authority, resource boundary and side-effect class are checked by the application. Define a narrow validated tool schema, restrict its identity permissions and require explicit authorization for irreversible or high-impact tool calls. The decisive evidence is that authorized read requests succeed; unapproved writes, cross-tenant identifiers and malformed arguments are rejected with traceable policy reasons.
Incorrect Answers
A team is investigating this task: A procurement agent may draft purchase orders, but any submission above a threshold requires verifiable approval. “The most powerful or least restricted model is always the safest choice regardless of task conditions” does not fully explain the decisive technical boundary in this case. A model-proposed tool call is untrusted until its arguments, caller authority, resource boundary and side-effect class are checked by the application.
“The fact that the team started work is enough to prove correctness for this exact case: a procurement agent may draft purchase orders, but any submission above a threshold requires verifiable approval” does not fully explain the decisive technical boundary in this case. A model-proposed tool call is untrusted until its arguments, caller authority, resource boundary and side-effect class are checked by the application. The implementation must define a narrow validated tool schema, restrict its identity permissions and require explicit authorization for irreversible or high-impact tool calls.
A procurement agent may draft purchase orders, but any submission above a threshold requires verifiable approval. “Successful provisioning by itself proves all security, output, and operational constraints are satisfied” does not fully explain the decisive technical boundary in this case. A model-proposed tool call is untrusted until its arguments, caller authority, resource boundary and side-effect class are checked by the application. The implementation must define a narrow validated tool schema, restrict its identity permissions and require explicit authorization for irreversible or high-impact tool calls.
Question 373
Architecture case: A procurement agent may draft purchase orders, but any submission above a threshold requires verifiable approval. What observable acceptance condition should be required before closing the task?
Explanation
Correct Answer
A model-proposed tool call is untrusted until its arguments, caller authority, resource boundary and side-effect class are checked by the application. Authorized read requests succeed; unapproved writes, cross-tenant identifiers and malformed arguments are rejected with traceable policy reasons. If this test fails, inspect function schema constraints, caller identity claims, effective tool permissions, validation errors and approval-event logs.
Incorrect Answers
“The operator reports success without reproducing the observed behavior: a procurement agent may draft purchase orders, but any submission above a threshold requires verifiable approval” alone does not establish all the required behavior. Authorized read requests succeed; unapproved writes, cross-tenant identifiers and malformed arguments are rejected with traceable policy reasons. A model-proposed tool call is untrusted until its arguments, caller authority, resource boundary and side-effect class are checked by the application.
Because a team is investigating this task: A procurement agent may draft purchase orders, but any submission above a threshold requires verifiable approval, “A project dashboard remains green, although the affected request, output and permissions were not checked” would not meet the requirement. Authorized read requests succeed; unapproved writes, cross-tenant identifiers and malformed arguments are rejected with traceable policy reasons.
A team is investigating this task: A procurement agent may draft purchase orders, but any submission above a threshold requires verifiable approval. “The team closes a change ticket without replaying a representative positive and negative test” alone does not establish all the required behavior. Authorized read requests succeed; unapproved writes, cross-tenant identifiers and malformed arguments are rejected with traceable policy reasons. A model-proposed tool call is untrusted until its arguments, caller authority, resource boundary and side-effect class are checked by the application.
Question 374
Production requirement: A procurement agent may draft purchase orders, but any submission above a threshold requires verifiable approval. Which first investigation is more useful than an unrelated health indicator?
Explanation
Correct Answer
A procurement agent may draft purchase orders, but any submission above a threshold requires verifiable approval. Inspect function schema constraints, caller identity claims, effective tool permissions, validation errors and approval-event logs. A model-proposed tool call is untrusted until its arguments, caller authority, resource boundary and side-effect class are checked by the application. Authorized read requests succeed; unapproved writes, cross-tenant identifiers and malformed arguments are rejected with traceable policy reasons.
Incorrect Answers
“Read only the broad service-health summary and skip actual workload evidence” would not meet the requirement. Inspect function schema constraints, caller identity claims, effective tool permissions, validation errors and approval-event logs. A model-proposed tool call is untrusted until its arguments, caller authority, resource boundary and side-effect class are checked by the application.
“Assume the general service is at fault without checking the affected operation in this case: a procurement agent may draft purchase orders, but any submission above a threshold requires verifiable approval” would not meet the requirement. Inspect function schema constraints, caller identity claims, effective tool permissions, validation errors and approval-event logs.
“Restart every affected component before collecting request or tool traces” would not be the most direct first check for this failure. Inspect function schema constraints, caller identity claims, effective tool permissions, validation errors and approval-event logs.
Question 375
A team is investigating this task: A procurement agent may draft purchase orders, but any submission above a threshold requires verifiable approval. Which TWO actions could conceal a serious failure while appearing faster? Select TWO answers.
Explanation
Correct Answer
Because a team is investigating this task: A procurement agent may draft purchase orders, but any submission above a threshold requires verifiable approval, “Execute any generated payment or deletion call without a human approval checkpoint” would violate the requirement. A model-proposed tool call is untrusted until its arguments, caller authority, resource boundary and side-effect class are checked by the application.
Because a team is investigating this task: A procurement agent may draft purchase orders, but any submission above a threshold requires verifiable approval, “Expose a general administrator tool to all customers because arguments are described in the prompt” would violate the requirement. A model-proposed tool call is untrusted until its arguments, caller authority, resource boundary and side-effect class are checked by the application.
Incorrect Answers
“Define a narrow validated tool schema, restrict its identity permissions and require explicit authorization for irreversible or high-impact tool calls” is the controlled implementation, not an unsafe workaround. A model-proposed tool call is untrusted until its arguments, caller authority, resource boundary and side-effect class are checked by the application. Confirm that authorized read requests succeed; unapproved writes, cross-tenant identifiers and malformed arguments are rejected with traceable policy reasons.
“Inspect records that explain the affected process, then check the result described in this case: A procurement agent may draft purchase orders” adds a relevant safeguard or verification step rather than weakening the solution. A model-proposed tool call is untrusted until its arguments, caller authority, resource boundary and side-effect class are checked by the application. Confirm that authorized read requests succeed; unapproved writes, cross-tenant identifiers and malformed arguments are rejected with traceable policy reasons.
“Authorized read requests succeed; unapproved writes, cross-tenant identifiers and malformed arguments are rejected with traceable policy reasons” is the required outcome of a working control, not an action that weakens it. A model-proposed tool call is untrusted until its arguments, caller authority, resource boundary and side-effect class are checked by the application.
Question 376
Situation: An HR search agent retrieves a policy page that includes a malicious instruction to reveal employee payroll records. Which implementation best meets the requirement while honoring the stated constraints?
Explanation
Correct Answer
Retrieved documents and tool outputs are data, not higher-priority instructions; application-side authorization must survive malicious text in the retrieved context. Confirm that known malicious retrieved snippets fail to redirect tool privileges or leak secrets while legitimate citation-based answers still succeed.
Incorrect Answers
“Apply the proposed setting to all connected workloads before testing the case-specific boundary: an hr search agent retrieves a policy page that includes a malicious instruction to reveal employee” does not implement the control that the facts call for. Retrieved documents and tool outputs are data, not higher-priority instructions; application-side authorization must survive malicious text in the retrieved context. Instead, separate retrieved untrusted content from system instructions, enforce tool and data policies outside the model, and test adversarial retrieval cases.
“Check only clean documents rather than adversarial input cases” does not implement the control that the facts call for. Retrieved documents and tool outputs are data, not higher-priority instructions; application-side authorization must survive malicious text in the retrieved context. Instead, separate retrieved untrusted content from system instructions, enforce tool and data policies outside the model, and test adversarial retrieval cases.
Because an HR search agent retrieves a policy page that includes a malicious instruction to reveal employee payroll records, “Promote every cited passage to a trusted instruction without verifying its source” does not implement the control that the facts call for. Retrieved documents and tool outputs are data, not higher-priority instructions; application-side authorization must survive malicious text in the retrieved context. Instead, separate retrieved untrusted content from system instructions, enforce tool and data policies outside the model, and test adversarial retrieval cases.
Question 377
Architecture case: An HR search agent retrieves a policy page that includes a malicious instruction to reveal employee payroll records. Which design boundary is most relevant to this decision?
Explanation
Correct Answer
An HR search agent retrieves a policy page that includes a malicious instruction to reveal employee payroll records. Retrieved documents and tool outputs are data, not higher-priority instructions; application-side authorization must survive malicious text in the retrieved context. Separate retrieved untrusted content from system instructions, enforce tool and data policies outside the model, and test adversarial retrieval cases.
Incorrect Answers
“The fact that the team started work is enough to prove correctness for this exact case: an hr search agent retrieves a policy page that includes a malicious instruction to reveal employee payroll records” does not fully explain the decisive technical boundary in this case. Retrieved documents and tool outputs are data, not higher-priority instructions; application-side authorization must survive malicious text in the retrieved context. The implementation must separate retrieved untrusted content from system instructions, enforce tool and data policies outside the model, and test adversarial retrieval cases.
“Successful provisioning by itself proves all security, output, and operational constraints are satisfied” does not fully explain the decisive technical boundary in this case. Retrieved documents and tool outputs are data, not higher-priority instructions; application-side authorization must survive malicious text in the retrieved context. The implementation must separate retrieved untrusted content from system instructions, enforce tool and data policies outside the model, and test adversarial retrieval cases.
An HR search agent retrieves a policy page that includes a malicious instruction to reveal employee payroll records. “The most powerful or least restricted model is always the safest choice regardless of task conditions” does not fully explain the decisive technical boundary in this case. Retrieved documents and tool outputs are data, not higher-priority instructions; application-side authorization must survive malicious text in the retrieved context. The implementation must separate retrieved untrusted content from system instructions, enforce tool and data policies outside the model, and test adversarial retrieval cases.
Question 378
Production requirement: An HR search agent retrieves a policy page that includes a malicious instruction to reveal employee payroll records. Which evidence is strongest for proving this requirement was met?
Explanation
Correct Answer
An HR search agent retrieves a policy page that includes a malicious instruction to reveal employee payroll records. Retrieved documents and tool outputs are data, not higher-priority instructions; application-side authorization must survive malicious text in the retrieved context. Known malicious retrieved snippets fail to redirect tool privileges or leak secrets while legitimate citation-based answers still succeed. If this test fails, inspect retrieval payload provenance, policy decision traces, attempted tool calls, rejected arguments and red-team evaluation outcomes.
Incorrect Answers
“The operator reports success without reproducing the observed behavior: an hr search agent retrieves a policy page that includes a malicious instruction to reveal employee payroll records” alone does not establish all the required behavior. Known malicious retrieved snippets fail to redirect tool privileges or leak secrets while legitimate citation-based answers still succeed. Retrieved documents and tool outputs are data, not higher-priority instructions; application-side authorization must survive malicious text in the retrieved context.
“A project dashboard remains green, although the affected request, output and permissions were not checked” would not meet the requirement. Known malicious retrieved snippets fail to redirect tool privileges or leak secrets while legitimate citation-based answers still succeed.
An HR search agent retrieves a policy page that includes a malicious instruction to reveal employee payroll records. “The team closes a change ticket without replaying a representative positive and negative test” alone does not establish all the required behavior. Known malicious retrieved snippets fail to redirect tool privileges or leak secrets while legitimate citation-based answers still succeed.
Question 379
A team is investigating this task: An HR search agent retrieves a policy page that includes a malicious instruction to reveal employee payroll records. What targeted records should the responder inspect to distinguish configuration from runtime behavior?
Explanation
Correct Answer
Inspect retrieval payload provenance, policy decision traces, attempted tool calls, rejected arguments and red-team evaluation outcomes. Retrieved documents and tool outputs are data, not higher-priority instructions; application-side authorization must survive malicious text in the retrieved context. Known malicious retrieved snippets fail to redirect tool privileges or leak secrets while legitimate citation-based answers still succeed.
Incorrect Answers
“Assume the general service is at fault without checking the affected operation in this case: an hr search agent retrieves a policy page that includes a malicious instruction to reveal employee payroll” would not meet the requirement. Inspect retrieval payload provenance, policy decision traces, attempted tool calls, rejected arguments and red-team evaluation outcomes.
Because an HR search agent retrieves a policy page that includes a malicious instruction to reveal employee payroll records, “Read only the broad service-health summary and skip actual workload evidence” would not meet the requirement. Inspect retrieval payload provenance, policy decision traces, attempted tool calls, rejected arguments and red-team evaluation outcomes. Retrieved documents and tool outputs are data, not higher-priority instructions; application-side authorization must survive malicious text in the retrieved context.
“Restart every affected component before collecting request or tool traces” would not be the most direct first check for this failure. Inspect retrieval payload provenance, policy decision traces, attempted tool calls, rejected arguments and red-team evaluation outcomes.
Question 380
Situation: An HR search agent retrieves a policy page that includes a malicious instruction to reveal employee payroll records. Which TWO proposed practices create material safety, correctness, or reliability risks? Select TWO answers.
Explanation
Correct Answer
An HR search agent retrieves a policy page that includes a malicious instruction to reveal employee payroll records. “Include production secret values in the retrieved document context to make the model more helpful” would violate the requirement. Retrieved documents and tool outputs are data, not higher-priority instructions; application-side authorization must survive malicious text in the retrieved context.
“Let a retrieved article override developer tool-access rules when it claims to be a system instruction” would violate the requirement. Retrieved documents and tool outputs are data, not higher-priority instructions; application-side authorization must survive malicious text in the retrieved context.
Incorrect Answers
Because an HR search agent retrieves a policy page that includes a malicious instruction to reveal employee payroll records, “Separate retrieved untrusted content from system instructions, enforce tool and data policies outside the model, and test adversarial retrieval cases” is the controlled implementation, not an unsafe workaround. Retrieved documents and tool outputs are data, not higher-priority instructions; application-side authorization must survive malicious text in the retrieved context. Confirm that known malicious retrieved snippets fail to redirect tool privileges or leak secrets while legitimate citation-based answers still succeed.
Because an HR search agent retrieves a policy page that includes a malicious instruction to reveal employee payroll records, “Known malicious retrieved snippets fail to redirect tool privileges or leak secrets while legitimate citation-based answers still succeed” is the required outcome of a working control, not an action that weakens it. Retrieved documents and tool outputs are data, not higher-priority instructions; application-side authorization must survive malicious text in the retrieved context. Confirm that known malicious retrieved snippets fail to redirect tool privileges or leak secrets while legitimate citation-based answers still succeed.
“Inspect records that explain the affected process, then check the result described in this case: An HR search agent retrieves a policy page that includes a malicious instruction to reveal employee” adds a relevant safeguard or verification step rather than weakening the solution. Retrieved documents and tool outputs are data, not higher-priority instructions; application-side authorization must survive malicious text in the retrieved context. Confirm that known malicious retrieved snippets fail to redirect tool privileges or leak secrets while legitimate citation-based answers still succeed.
Developing AI Apps and Agents on Azure (AI-103) Practice Tests
- Microsoft AI-103 Practice Test 36 - 10 Questions (Last Reviewed 17-Oct-2026)
- Microsoft AI-103 Practice Test 37 - 10 Questions (Last Reviewed 16-Oct-2026)
- Microsoft AI-103 Practice Test 38 - 10 Questions (Last Reviewed 15-Oct-2026)
- Microsoft AI-103 Practice Test 39 - 10 Questions (Last Reviewed 14-Oct-2026)
- Microsoft AI-103 Practice Test 40 - 10 Questions (Last Reviewed 13-Oct-2026)