Question 110
A solution architect is deciding how to explain changes in moderation behavior across releases and support rollback. Which Azure AI pattern addresses that task?
- Limit knowledge access to the caller's entitlements
- Apply input and output harm filters
- Separate review and execution responsibilities
- Enforce per-tool authorization independently
- Use documented safety-policy versioning
Explanation
Correct Answer
E:For the stated requirement to explain changes in moderation behavior across releases and support rollback, versioned guardrail policies make thresholds, blocklists and exception decisions reproducible in evaluations and incidents; this is the mechanism that the system must implement.
Incorrect Answers
A:For a system whose objective is to prevent an agent from citing restricted HR content to ordinary staff, security trimming or permission-aware retrieval applies user access rules before private passages reach generation; this does not directly meet the required outcome to explain changes in moderation behavior across releases and support rollback.
B:For a system whose objective is to limit harmful user prompts and unsafe assistant responses according to an approved safety policy, content filters classify supported harm categories at the inference boundary; thresholds require evaluation for false positives and misses; this does not directly meet the required outcome to explain changes in moderation behavior across releases and support rollback.
C:For a system whose objective is to stop an agent from approving its own high-risk recommendation, distinct approval identity or review service introduces independent oversight; a second prompt in the same privileged agent is not separation of duties; this does not directly meet the required outcome to explain changes in moderation behavior across releases and support rollback.
D:For a system whose objective is to stop an agent's broad language permissions from turning into unrestricted business-system privileges, tools must authorize each action using the caller's allowed scope, regardless of what the LLM requests; natural-language rules are insufficient; this does not directly meet the required outcome to explain changes in moderation behavior across releases and support rollback.
Question 111
The implementation does not meet its business requirement: Support traces contain unredacted customer IDs and telephone numbers that the monitoring team does not need. Which corrective action is appropriate?
- Put a checked approval state between proposal and execution for payment-changing tools
- Enable appropriate indirect-injection screening and isolate retrieved text as untrusted data
- Redact sensitive fields at collection and restrict trace access according to retention policy
- Persist source-version identifiers and retrieval metadata with each audited decision
Explanation
Correct Answer
C:The fault signal is that support traces contain unredacted customer IDs and telephone numbers that the monitoring team does not need; the targeted fix works because PII-aware redaction and minimal logging reduce privacy risk; collecting full prompts in diagnostic systems is not inherently necessary, restoring the ability to keep traces useful without retaining customer identifiers or secrets in operational logs.
Incorrect Answers
A:If the failing function were to keep payment or account-changing tool calls under an explicit human authorization step, this change could be relevant because human-in-the-loop approval gates restrict irreversible operations even if the agent's reasoning or external evidence is wrong; the actual symptom is that support traces contain unredacted customer IDs and telephone numbers that the monitoring team does not need, which calls for a different repair.
B:If the failing function were to prevent an untrusted document from instructing an agent to ignore its governing rules, this change could be relevant because prompt Shields and related controls can detect indirect instructions embedded in lower-trust content; document text remains evidence, not authority; the actual symptom is that support traces contain unredacted customer IDs and telephone numbers that the monitoring team does not need, which calls for a different repair.
D:If the failing function were to show exactly which source version supported a regulated answer, this change could be relevant because document IDs, timestamps, and citation spans let reviewers reconstruct the evidence used at response time; later source edits should not erase the audit trail; the actual symptom is that support traces contain unredacted customer IDs and telephone numbers that the monitoring team does not need, which calls for a different repair.
Question 112
Stakeholders want a Foundry-based application able to keep payment or account-changing tool calls under an explicit human authorization step. Which choice implements that requirement?
- Separate review and execution responsibilities
- Remove or mask sensitive personal data before logging
- Require human approval for high-impact agent actions
- Apply input and output harm filters
Explanation
Correct Answer
C:For the stated requirement to keep payment or account-changing tool calls under an explicit human authorization step, human-in-the-loop approval gates restrict irreversible operations even if the agent's reasoning or external evidence is wrong; this is the mechanism that the system must implement.
Incorrect Answers
A:For a system whose objective is to stop an agent from approving its own high-risk recommendation, distinct approval identity or review service introduces independent oversight; a second prompt in the same privileged agent is not separation of duties; this does not directly meet the required outcome to keep payment or account-changing tool calls under an explicit human authorization step.
B:For a system whose objective is to keep traces useful without retaining customer identifiers or secrets in operational logs, PII-aware redaction and minimal logging reduce privacy risk; collecting full prompts in diagnostic systems is not inherently necessary; this does not directly meet the required outcome to keep payment or account-changing tool calls under an explicit human authorization step.
D:For a system whose objective is to limit harmful user prompts and unsafe assistant responses according to an approved safety policy, content filters classify supported harm categories at the inference boundary; thresholds require evaluation for false positives and misses; this does not directly meet the required outcome to keep payment or account-changing tool calls under an explicit human authorization step.
Question 113
An engineer receives this incident report: A public help agent retrieves confidential salary guidance because it uses a shared unrestricted index credential. What should be changed?
- Require approval by a separate authorized reviewer before committing regulated outcomes
- Version the safety configuration with release artifacts and preserve change history
- Redact sensitive fields at collection and restrict trace access according to retention policy
- Filter retrieval to documents authorized for the current caller and enforce access at the source
Explanation
Correct Answer
D:The fault signal is that a public help agent retrieves confidential salary guidance because it uses a shared unrestricted index credential; the targeted fix works because security trimming or permission-aware retrieval applies user access rules before private passages reach generation, restoring the ability to prevent an agent from citing restricted HR content to ordinary staff.
Incorrect Answers
A:If the failing function were to stop an agent from approving its own high-risk recommendation, this change could be relevant because distinct approval identity or review service introduces independent oversight; a second prompt in the same privileged agent is not separation of duties; the actual symptom is that a public help agent retrieves confidential salary guidance because it uses a shared unrestricted index credential, which calls for a different repair.
B:If the failing function were to explain changes in moderation behavior across releases and support rollback, this change could be relevant because versioned guardrail policies make thresholds, blocklists and exception decisions reproducible in evaluations and incidents; the actual symptom is that a public help agent retrieves confidential salary guidance because it uses a shared unrestricted index credential, which calls for a different repair.
C:If the failing function were to keep traces useful without retaining customer identifiers or secrets in operational logs, this change could be relevant because PII-aware redaction and minimal logging reduce privacy risk; collecting full prompts in diagnostic systems is not inherently necessary; the actual symptom is that a public help agent retrieves confidential salary guidance because it uses a shared unrestricted index credential, which calls for a different repair.
Question 114
Engineers have implemented a change to record provenance for retrieved evidence. What validation would expose a remaining defect?
- Attempt denied write calls and confirm enforced rejection plus audit records
- Audit that proposer and approver identities are distinct on completed decisions
- Evaluate precision and recall for the affected safety categories after adjustment
- Reconstruct historical answer citations using retained source references and version records
Explanation
Correct Answer
D:The accepted change is to record provenance for retrieved evidence; this evaluation checks whether it can show exactly which source version supported a regulated answer, since document IDs, timestamps, and citation spans let reviewers reconstruct the evidence used at response time; later source edits should not erase the audit trail.
Incorrect Answers
A:This test concerns the ability to stop an agent's broad language permissions from turning into unrestricted business-system privileges, for which tools must authorize each action using the caller's allowed scope, regardless of what the LLM requests; natural-language rules are insufficient; it does not establish whether the change to record provenance for retrieved evidence works for the current release.
B:This test concerns the ability to stop an agent from approving its own high-risk recommendation, for which distinct approval identity or review service introduces independent oversight; a second prompt in the same privileged agent is not separation of duties; it does not establish whether the change to record provenance for retrieved evidence works for the current release.
C:This test concerns the ability to avoid treating all newly flagged legitimate customer requests as genuine safety threats, for which human-reviewed labeled examples distinguish benign business terms from actual prohibited content; thresholds and policy need measured tuning; it does not establish whether the change to record provenance for retrieved evidence works for the current release.
Question 115
Production telemetry reveals this failure: An agent repeatedly calls the same search tool without making progress and its token spending spikes. Which change most directly corrects it?
- Version the safety configuration with release artifacts and preserve change history
- Require approval by a separate authorized reviewer before committing regulated outcomes
- Set a maximum step count, tool-call budget, and explicit abort or handoff rules
- Add attack-oriented scenarios for jailbreaks, data leakage, and unsafe tool use before release
Explanation
Correct Answer
C:The fault signal is that an agent repeatedly calls the same search tool without making progress and its token spending spikes; the targeted fix works because execution limits on tool calls, runtime and cost bound what autonomy can do, independent of prompt compliance, restoring the ability to prevent infinite tool loops and uncontrolled consumption in long-running workflows.
Incorrect Answers
A:If the failing function were to explain changes in moderation behavior across releases and support rollback, this change could be relevant because versioned guardrail policies make thresholds, blocklists and exception decisions reproducible in evaluations and incidents; the actual symptom is that an agent repeatedly calls the same search tool without making progress and its token spending spikes, which calls for a different repair.
B:If the failing function were to stop an agent from approving its own high-risk recommendation, this change could be relevant because distinct approval identity or review service introduces independent oversight; a second prompt in the same privileged agent is not separation of duties; the actual symptom is that an agent repeatedly calls the same search tool without making progress and its token spending spikes, which calls for a different repair.
D:If the failing function were to discover harmful outputs and instruction-bypass failures across diverse user and tool inputs, this change could be relevant because red-team cases exercise guardrail boundaries under realistic attacks, unlike happy-path functional tests alone; the actual symptom is that an agent repeatedly calls the same search tool without making progress and its token spending spikes, which calls for a different repair.
Question 116
Stakeholders want a production AI solution able to limit harmful user prompts and unsafe assistant responses according to an approved safety policy. Which choice implements that requirement?
- Remove or mask sensitive personal data before logging
- Apply input and output harm filters
- Use prompt-injection detection on retrieved content
- Enforce per-tool authorization independently
Explanation
Correct Answer
B:For the stated requirement to limit harmful user prompts and unsafe assistant responses according to an approved safety policy, content filters classify supported harm categories at the inference boundary; thresholds require evaluation for false positives and misses; this is the mechanism that the system must implement.
Incorrect Answers
A:For a system whose objective is to keep traces useful without retaining customer identifiers or secrets in operational logs, PII-aware redaction and minimal logging reduce privacy risk; collecting full prompts in diagnostic systems is not inherently necessary; this does not directly meet the required outcome to limit harmful user prompts and unsafe assistant responses according to an approved safety policy.
C:For a system whose objective is to prevent an untrusted document from instructing an agent to ignore its governing rules, prompt Shields and related controls can detect indirect instructions embedded in lower-trust content; document text remains evidence, not authority; this does not directly meet the required outcome to limit harmful user prompts and unsafe assistant responses according to an approved safety policy.
D:For a system whose objective is to stop an agent's broad language permissions from turning into unrestricted business-system privileges, tools must authorize each action using the caller's allowed scope, regardless of what the LLM requests; natural-language rules are insufficient; this does not directly meet the required outcome to limit harmful user prompts and unsafe assistant responses according to an approved safety policy.
Question 117
A design review proposes to require human approval for high-impact agent actions. Which measurable check would establish success on real inputs?
- Sample logs to verify redaction coverage and authorized retention behavior
- Test cross-role queries for unauthorized citations and verify least-privilege results
- Test that execution is blocked without approval and that every approval is attributable
- Red-team malicious instructions inside retrieved pages and record detected unsafe tool intentions
Explanation
Correct Answer
C:The accepted change is to require human approval for high-impact agent actions; this evaluation checks whether it can keep payment or account-changing tool calls under an explicit human authorization step, since human-in-the-loop approval gates restrict irreversible operations even if the agent's reasoning or external evidence is wrong.
Incorrect Answers
A:This test concerns the ability to keep traces useful without retaining customer identifiers or secrets in operational logs, for which PII-aware redaction and minimal logging reduce privacy risk; collecting full prompts in diagnostic systems is not inherently necessary; it does not establish whether the change to require human approval for high-impact agent actions works for the current release.
B:This test concerns the ability to prevent an agent from citing restricted HR content to ordinary staff, for which security trimming or permission-aware retrieval applies user access rules before private passages reach generation; it does not establish whether the change to require human approval for high-impact agent actions works for the current release.
D:This test concerns the ability to prevent an untrusted document from instructing an agent to ignore its governing rules, for which prompt Shields and related controls can detect indirect instructions embedded in lower-trust content; document text remains evidence, not authority; it does not establish whether the change to require human approval for high-impact agent actions works for the current release.
Question 118
A new Foundry-based application must prevent an untrusted document from instructing an agent to ignore its governing rules. Which approach should the team select first?
- Limit knowledge access to the caller's entitlements
- Use prompt-injection detection on retrieved content
- Apply input and output harm filters
- Require human approval for high-impact agent actions
Explanation
Correct Answer
B:For the stated requirement to prevent an untrusted document from instructing an agent to ignore its governing rules, prompt Shields and related controls can detect indirect instructions embedded in lower-trust content; document text remains evidence, not authority; this is the mechanism that the system must implement.
Incorrect Answers
A:For a system whose objective is to prevent an agent from citing restricted HR content to ordinary staff, security trimming or permission-aware retrieval applies user access rules before private passages reach generation; this does not directly meet the required outcome to prevent an untrusted document from instructing an agent to ignore its governing rules.
C:For a system whose objective is to limit harmful user prompts and unsafe assistant responses according to an approved safety policy, content filters classify supported harm categories at the inference boundary; thresholds require evaluation for false positives and misses; this does not directly meet the required outcome to prevent an untrusted document from instructing an agent to ignore its governing rules.
D:For a system whose objective is to keep payment or account-changing tool calls under an explicit human authorization step, human-in-the-loop approval gates restrict irreversible operations even if the agent's reasoning or external evidence is wrong; this does not directly meet the required outcome to prevent an untrusted document from instructing an agent to ignore its governing rules.
Question 119
In an architecture review, the key requirement is to show exactly which source version supported a regulated answer. Which solution directly satisfies it?
- Enforce per-tool authorization independently
- Remove or mask sensitive personal data before logging
- Review false positives after moderation updates
- Record provenance for retrieved evidence
Explanation
Correct Answer
D:For the stated requirement to show exactly which source version supported a regulated answer, document IDs, timestamps, and citation spans let reviewers reconstruct the evidence used at response time; later source edits should not erase the audit trail; this is the mechanism that the system must implement.
Incorrect Answers
A:For a system whose objective is to stop an agent's broad language permissions from turning into unrestricted business-system privileges, tools must authorize each action using the caller's allowed scope, regardless of what the LLM requests; natural-language rules are insufficient; this does not directly meet the required outcome to show exactly which source version supported a regulated answer.
B:For a system whose objective is to keep traces useful without retaining customer identifiers or secrets in operational logs, PII-aware redaction and minimal logging reduce privacy risk; collecting full prompts in diagnostic systems is not inherently necessary; this does not directly meet the required outcome to show exactly which source version supported a regulated answer.
C:For a system whose objective is to avoid treating all newly flagged legitimate customer requests as genuine safety threats, human-reviewed labeled examples distinguish benign business terms from actual prohibited content; thresholds and policy need measured tuning; this does not directly meet the required outcome to show exactly which source version supported a regulated answer.
Question 120
A new customer-facing AI service must prevent an agent from citing restricted HR content to ordinary staff. Which approach should the team select first?
- Constrain an autonomous agent's steps and budget
- Limit knowledge access to the caller's entitlements
- Use prompt-injection detection on retrieved content
- Separate review and execution responsibilities
Explanation
Correct Answer
B:For the stated requirement to prevent an agent from citing restricted HR content to ordinary staff, security trimming or permission-aware retrieval applies user access rules before private passages reach generation; this is the mechanism that the system must implement.
Incorrect Answers
A:For a system whose objective is to prevent infinite tool loops and uncontrolled consumption in long-running workflows, execution limits on tool calls, runtime and cost bound what autonomy can do, independent of prompt compliance; this does not directly meet the required outcome to prevent an agent from citing restricted HR content to ordinary staff.
C:For a system whose objective is to prevent an untrusted document from instructing an agent to ignore its governing rules, prompt Shields and related controls can detect indirect instructions embedded in lower-trust content; document text remains evidence, not authority; this does not directly meet the required outcome to prevent an agent from citing restricted HR content to ordinary staff.
D:For a system whose objective is to stop an agent from approving its own high-risk recommendation, distinct approval identity or review service introduces independent oversight; a second prompt in the same privileged agent is not separation of duties; this does not directly meet the required outcome to prevent an agent from citing restricted HR content to ordinary staff.