Microsoft AI-103 Practice Test 39 - Prepaway
Practice Exams:

Pass Microsoft AI-103 Exam in First Attempt Guaranteed!

Get 100% Latest Exam Questions, Accurate & Verified Answers!

30 Days Free Updates, Instant Download!

Verified By Experts
Premium File
$89.99 $98.99

AI-103 Premium ETE File


  • Premium File 187 Questions & Answers. Last Update: Nov 15, 2026
Whats Included:
  • Latest Questions
  • 100% Accurate Answers
  • Fast Exam Updates

Microsoft AI-103 Practice Test 39

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.

  1. Include production secret values in the retrieved document context to make the model more helpful
  2. Separate retrieved untrusted content from system instructions, enforce tool and data policies outside the model, and test adversarial retrieval cases
  3. Known malicious retrieved snippets fail to redirect tool privileges or leak secrets while legitimate citation-based answers still succeed
  4. 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 payroll records
  5. Let a retrieved article override developer tool-access rules when it claims to be a system instruction

Explanation

Correct Answer

A:

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.

E:

“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

B:

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.

C:

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.

D:

“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.

Question 381

Architecture case: A support assistant encounters a vendor README that asks the agent to send environment secrets to an external URL. What configuration should the engineer choose for this workload?

  1. Promote every cited passage to a trusted instruction without verifying its source
  2. Check only clean documents rather than adversarial input cases
  3. Separate retrieved untrusted content from system instructions, enforce tool and data policies outside the model, and test adversarial retrieval cases
  4. Apply the proposed setting to all connected workloads before testing the case-specific boundary: a support assistant encounters a vendor readme that asks the agent to send environment secrets to an external url

Explanation

Correct Answer

C:

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. If the result fails, inspect retrieval payload provenance, policy decision traces, attempted tool calls, rejected arguments and red-team evaluation outcomes.

Incorrect Answers

A:

Because a support assistant encounters a vendor README that asks the agent to send environment secrets to an external URL, “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.

B:

Because a support assistant encounters a vendor README that asks the agent to send environment secrets to an external URL, “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.

D:

“Apply the proposed setting to all connected workloads before testing the case-specific boundary: a support assistant encounters a vendor readme that asks the agent to send environment secrets to an” 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 382

Production requirement: A support assistant encounters a vendor README that asks the agent to send environment secrets to an external URL. What technical fact prevents the simplest shortcut from being reliable?

  1. The fact that the team started work is enough to prove correctness for this exact case: a support assistant encounters a vendor readme that asks the agent to send environment secrets to an external url
  2. Retrieved documents and tool outputs are data, not higher-priority instructions; application-side authorization must survive malicious text in the retrieved context
  3. The most powerful or least restricted model is always the safest choice regardless of task conditions
  4. Successful provisioning by itself proves all security, output, and operational constraints are satisfied

Explanation

Correct Answer

B:

Because a support assistant encounters a vendor README that asks the agent to send environment secrets to an external URL, 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

A:

“The fact that the team started work is enough to prove correctness for this exact case: a support assistant encounters a vendor readme that asks the agent to send environment secrets to an external url” 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.

C:

“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.

D:

A support assistant encounters a vendor README that asks the agent to send environment secrets to an external URL. “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.

Question 383

A team is investigating this task: A support assistant encounters a vendor README that asks the agent to send environment secrets to an external URL. What end-to-end test would distinguish successful operation from an installed configuration?

  1. The team closes a change ticket without replaying a representative positive and negative test
  2. Known malicious retrieved snippets fail to redirect tool privileges or leak secrets while legitimate citation-based answers still succeed
  3. A project dashboard remains green, although the affected request, output and permissions were not checked
  4. The operator reports success without reproducing the observed behavior: a support assistant encounters a vendor readme that asks the agent to send environment secrets to an external url

Explanation

Correct Answer

B:

Because a support assistant encounters a vendor README that asks the agent to send environment secrets to an external URL, 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

A:

A support assistant encounters a vendor README that asks the agent to send environment secrets to an external URL. “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.

C:

Because a support assistant encounters a vendor README that asks the agent to send environment secrets to an external URL, “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.

D:

“The operator reports success without reproducing the observed behavior: a support assistant encounters a vendor readme that asks the agent to send environment secrets to an external url” 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 384

Situation: A support assistant encounters a vendor README that asks the agent to send environment secrets to an external URL. Which diagnostic investigation best isolates the issue before another change?

  1. Read only the broad service-health summary and skip actual workload evidence
  2. Restart every affected component before collecting request or tool traces
  3. Inspect retrieval payload provenance, policy decision traces, attempted tool calls, rejected arguments and red-team evaluation outcomes
  4. Assume the general service is at fault without checking the affected operation in this case: a support assistant encounters a vendor readme that asks the agent to send environment secrets to an external url

Explanation

Correct Answer

C:

A support assistant encounters a vendor README that asks the agent to send environment secrets to an external URL. 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

A:

A support assistant encounters a vendor README that asks the agent to send environment secrets to an external URL. “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.

B:

Because a support assistant encounters a vendor README that asks the agent to send environment secrets to an external URL, “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. Retrieved documents and tool outputs are data, not higher-priority instructions; application-side authorization must survive malicious text in the retrieved context.

D:

“Assume the general service is at fault without checking the affected operation in this case: a support assistant encounters a vendor readme that asks the agent to send environment secrets to an external” 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.

Question 385

Architecture case: A support assistant encounters a vendor README that asks the agent to send environment secrets to an external URL. Which TWO shortcuts should be rejected for this scenario? Select TWO answers.

  1. Include production secret values in the retrieved document context to make the model more helpful
  2. Inspect records that explain the affected process, then check the result described in this case: A support assistant encounters a vendor README that asks the agent to send environment secrets to an external URL
  3. Let a retrieved article override developer tool-access rules when it claims to be a system instruction
  4. Known malicious retrieved snippets fail to redirect tool privileges or leak secrets while legitimate citation-based answers still succeed
  5. Separate retrieved untrusted content from system instructions, enforce tool and data policies outside the model, and test adversarial retrieval cases

Explanation

Correct Answer

A:

“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.

C:

A support assistant encounters a vendor README that asks the agent to send environment secrets to an external URL. “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

B:

“Inspect records that explain the affected process, then check the result described in this case: A support assistant encounters a vendor README that asks the agent to send environment secrets to an” 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.

D:

A support assistant encounters a vendor README that asks the agent to send environment secrets to an external URL. “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.

E:

Because a support assistant encounters a vendor README that asks the agent to send environment secrets to an external URL, “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.

Question 386

Production requirement: A financial RAG tool retrieves an external PDF containing text impersonating a new administrator message. Which next change most directly addresses the reported technical risk?

  1. Check only clean documents rather than adversarial input cases
  2. Apply the proposed setting to all connected workloads before testing the case-specific boundary: a financial rag tool retrieves an external pdf containing text impersonating a new administrator message
  3. Promote every cited passage to a trusted instruction without verifying its source
  4. Separate retrieved untrusted content from system instructions, enforce tool and data policies outside the model, and test adversarial retrieval cases

Explanation

Correct Answer

D:

Separate retrieved untrusted content from system instructions, enforce tool and data policies outside the model, and test adversarial retrieval cases. Confirm that known malicious retrieved snippets fail to redirect tool privileges or leak secrets while legitimate citation-based answers still succeed. If the result fails, retrieved documents and tool outputs are data, not higher-priority instructions; application-side authorization must survive malicious text in the retrieved context.

Incorrect Answers

A:

“Check only clean documents rather than adversarial input cases” does not implement the control that the facts call for. Instead, separate retrieved untrusted content from system instructions, enforce tool and data policies outside the model, and test adversarial retrieval cases.

B:

“Apply the proposed setting to all connected workloads before testing the case-specific boundary: a financial rag tool retrieves an external pdf containing text impersonating a new administrator message” does not implement the control that the facts call for. Instead, separate retrieved untrusted content from system instructions, enforce tool and data policies outside the model, and test adversarial retrieval cases.

C:

“Promote every cited passage to a trusted instruction without verifying its source” does not implement the control that the facts call for. Instead, separate retrieved untrusted content from system instructions, enforce tool and data policies outside the model, and test adversarial retrieval cases.

Question 387

A team is investigating this task: A financial RAG tool retrieves an external PDF containing text impersonating a new administrator message. Which explanation best captures the key behavior of this architecture?

  1. The most powerful or least restricted model is always the safest choice regardless of task conditions
  2. Retrieved documents and tool outputs are data, not higher-priority instructions; application-side authorization must survive malicious text in the retrieved context
  3. Successful provisioning by itself proves all security, output, and operational constraints are satisfied
  4. The fact that the team started work is enough to prove correctness for this exact case: a financial rag tool retrieves an external pdf containing text impersonating a new administrator message

Explanation

Correct Answer

B:

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

A:

“The most powerful or least restricted model is always the safest choice regardless of task conditions” would not be the most direct first check for this failure. Retrieved documents and tool outputs are data, not higher-priority instructions; application-side authorization must survive malicious text in the retrieved context.

C:

“Successful provisioning by itself proves all security, output, and operational constraints are satisfied” would not be the most direct first check for this failure. Retrieved documents and tool outputs are data, not higher-priority instructions; application-side authorization must survive malicious text in the retrieved context.

D:

“The fact that the team started work is enough to prove correctness for this exact case: a financial rag tool retrieves an external pdf containing text impersonating a new administrator message” would not be the most direct first check for this failure. Retrieved documents and tool outputs are data, not higher-priority instructions; application-side authorization must survive malicious text in the retrieved context.

Question 388

Situation: A financial RAG tool retrieves an external PDF containing text impersonating a new administrator message. Which result would demonstrate that the proposed remedy works end to end?

  1. Known malicious retrieved snippets fail to redirect tool privileges or leak secrets while legitimate citation-based answers still succeed
  2. A project dashboard remains green, although the affected request, output and permissions were not checked
  3. The team closes a change ticket without replaying a representative positive and negative test
  4. The operator reports success without reproducing the observed behavior: a financial rag tool retrieves an external pdf containing text impersonating a new administrator message

Explanation

Correct Answer

A:

Because a financial RAG tool retrieves an external PDF containing text impersonating a new administrator message, known malicious retrieved snippets fail to redirect tool privileges or leak secrets while legitimate citation-based answers still succeed. Separate retrieved untrusted content from system instructions, enforce tool and data policies outside the model, and test adversarial retrieval cases.

Incorrect Answers

B:

Because a financial RAG tool retrieves an external PDF containing text impersonating a new administrator message, “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.

C:

Because a financial RAG tool retrieves an external PDF containing text impersonating a new administrator message, “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.

D:

“The operator reports success without reproducing the observed behavior: a financial rag tool retrieves an external pdf containing text impersonating a new administrator message” 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 389

Architecture case: A financial RAG tool retrieves an external PDF containing text impersonating a new administrator message. What evidence should the engineer correlate to verify the actual failure cause?

  1. Inspect retrieval payload provenance, policy decision traces, attempted tool calls, rejected arguments and red-team evaluation outcomes
  2. Restart every affected component before collecting request or tool traces
  3. Assume the general service is at fault without checking the affected operation in this case: a financial rag tool retrieves an external pdf containing text impersonating a new administrator message
  4. Read only the broad service-health summary and skip actual workload evidence

Explanation

Correct Answer

A:

Because a financial RAG tool retrieves an external PDF containing text impersonating a new administrator message, known malicious retrieved snippets fail to redirect tool privileges or leak secrets while legitimate citation-based answers still succeed. Separate retrieved untrusted content from system instructions, enforce tool and data policies outside the model, and test adversarial retrieval cases. The observable acceptance condition here is: Inspect retrieval payload provenance, policy decision traces, attempted tool calls, rejected arguments and red-team evaluation outcomes.

Incorrect Answers

B:

“Restart every affected component before collecting request or tool traces” 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.

C:

“Assume the general service is at fault without checking the affected operation in this case: a financial rag tool retrieves an external pdf containing text impersonating a new administrator message” would not meet the requirement. Known malicious retrieved snippets fail to redirect tool privileges or leak secrets while legitimate citation-based answers still succeed.

D:

“Read only the broad service-health summary and skip actual workload evidence” would not meet the requirement. Known malicious retrieved snippets fail to redirect tool privileges or leak secrets while legitimate citation-based answers still succeed.

Question 390

Production requirement: A financial RAG tool retrieves an external PDF containing text impersonating a new administrator message. Which TWO decisions would undermine the required system boundary? Select TWO answers.

  1. Separate retrieved untrusted content from system instructions, enforce tool and data policies outside the model, and test adversarial retrieval cases
  2. Include production secret values in the retrieved document context to make the model more helpful
  3. Let a retrieved article override developer tool-access rules when it claims to be a system instruction
  4. Known malicious retrieved snippets fail to redirect tool privileges or leak secrets while legitimate citation-based answers still succeed
  5. Inspect records that explain the affected process, then check the result described in this case: A financial RAG tool retrieves an external PDF containing text impersonating a new administrator message

Explanation

Correct Answer

B:

“Include production secret values in the retrieved document context to make the model more helpful” would violate the requirement.

C:

“Let a retrieved article override developer tool-access rules when it claims to be a system instruction” would violate the requirement.

Incorrect Answers

A:

“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. Confirm that known malicious retrieved snippets fail to redirect tool privileges or leak secrets while legitimate citation-based answers still succeed.

D:

“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. Confirm that known malicious retrieved snippets fail to redirect tool privileges or leak secrets while legitimate citation-based answers still succeed.

E:

“Inspect records that explain the affected process, then check the result described in this case: A financial RAG tool retrieves an external PDF containing text impersonating a new administrator message” adds a relevant safeguard or verification step rather than weakening the solution.

Developing AI Apps and Agents on Azure (AI-103) Practice Tests