Microsoft AB-100 Dumps
Microsoft AB-100 Practice Test 50
Question 490
During an AI adoption planning, an AI governance lead must accept a change for prompt library regression. The issue to be addressed is: A team changes a shared prompt instruction and unrelated customer-service answers become unsafe. Which concrete verification action demonstrates that the relevant behavior works?
Explanation
Correct Answer
To verify this change, the research institute should compare responses across the previous and proposed prompt on protected cases. Operators still need the implementation to run versioned intent and policy regression cases whenever a governed prompt changes and the preventive requirement to require change-linked regression gates for every centrally managed prompt.
Incorrect Answers
The action to run versioned intent and policy regression cases whenever a governed prompt changes can contribute to prompt library regression as a implementation step; here, a team changes a shared prompt instruction and unrelated customer-service answers become unsafe. The more directly responsive choice is to compare responses across the previous and proposed prompt on protected cases.
The assumption to test only one favorite question after modifying a shared production prompt is the exact failure pattern the team needs to avoid. The relevant change is to run versioned intent and policy regression cases whenever a governed prompt changes; a credible check must compare responses across the previous and proposed prompt on protected cases.
For agent evaluation set representativeness, it can be sensible to compare evaluation coverage with the top production intents and known incidents if an agent is evaluated only on easy happy-path prompts despite frequent ambiguous production requests. Without the actual change to compare responses across the previous and proposed prompt on protected cases, the answer lacks the correct operational scope even if the alternate action is technically valid elsewhere.
Question 491
At a logistics enterprise, the AI program manager is preparing a customer service escalation review. The concrete requirement is: A retrieved web page tells an agent to ignore its instructions and leak the current customer record. Which design action BEST addresses the stated need?
Explanation
Correct Answer
The proposed implementation is to include prompt-manipulation scenarios in predeployment tool-permission and retrieval testing; this matches the operational requirement that a retrieved web page tells an agent to ignore its instructions and leak the current customer record. The team can inject a fake system instruction in a retrieved page and confirm data egress is denied. The team should also test tool boundaries against untrusted instructions and independent policy guards.
Incorrect Answers
Choosing to test interruptions turn taking and duplicate side effects on real voice paths would direct effort toward voice-agent interruption test, whose characteristic issue is that a voice-based service agent restarts a payment confirmation whenever the customer interrupts. The relevant corrective or preventive action here remains to include prompt-manipulation scenarios in predeployment tool-permission and retrieval testing, with its own validation criteria.
The assumption to treat malicious webpage content as a trusted enterprise system command is the exact failure pattern the team needs to avoid. The relevant change is to include prompt-manipulation scenarios in predeployment tool-permission and retrieval testing; a credible check must inject a fake system instruction in a retrieved page and confirm data egress is denied.
The action to inject a fake system instruction in a retrieved page and confirm data egress is denied can contribute to adversarial instruction testing as a targeted acceptance check step; here, a retrieved web page tells an agent to ignore its instructions and leak the current customer record. The more directly responsive choice is to include prompt-manipulation scenarios in predeployment tool-permission and retrieval testing.
Question 492
After an agent reliability assessment at a pharmaceutical supplier, the team observes this failure: A retrieved web page tells an agent to ignore its instructions and leak the current customer record. Which flawed assumption or implementation choice would MOST DIRECTLY explain why the intended outcome was not achieved?
Explanation
Correct Answer
This design mistake is to treat malicious webpage content as a trusted enterprise system command; it conflicts with the actual adversarial instruction testing behavior needed when a retrieved web page tells an agent to ignore its instructions and leak the current customer record. The pharmaceutical supplier should instead include prompt-manipulation scenarios in predeployment tool-permission and retrieval testing, followed by a check to inject a fake system instruction in a retrieved page and confirm data egress is denied.
Incorrect Answers
The option to pass the launch gate using only the same five demonstrations used during development would normally serve a different agent evaluation set representativeness requirement, where an agent is evaluated only on easy happy-path prompts despite frequent ambiguous production requests. The alternative may be useful in its own context, but it does not directly supply treat malicious webpage content as a trusted enterprise system command for the stated decision.
The alternative to include prompt-manipulation scenarios in predeployment tool-permission and retrieval testing is a implementation step rather than the faulty assumption being diagnosed. The failure in adversarial instruction testing arises from treat malicious webpage content as a trusted enterprise system command; the concrete symptom is that a retrieved web page tells an agent to ignore its instructions and leak the current customer record.
The alternative to inject a fake system instruction in a retrieved page and confirm data egress is denied is a targeted acceptance check step rather than the faulty assumption being diagnosed. The failure in adversarial instruction testing arises from treat malicious webpage content as a trusted enterprise system command; the concrete symptom is that a retrieved web page tells an agent to ignore its instructions and leak the current customer record.
Question 493
The regional hospital group is responding to a preproduction release gate. An integration engineer discovers this issue: A retrieved web page tells an agent to ignore its instructions and leak the current customer record. What corrective step should the team take to restore the intended functionality most directly?
Explanation
Correct Answer
The direct repair is to block the attempted exfiltration and isolate the injected source. Once the condition is corrected, the regional hospital group should inject a fake system instruction in a retrieved page and confirm data egress is denied. To reduce recurrence, it should also test tool boundaries against untrusted instructions and independent policy guards; those steps provide different recovery and prevention functions.
Incorrect Answers
The assumption to treat malicious webpage content as a trusted enterprise system command is the exact failure pattern the team needs to avoid. The relevant change is to include prompt-manipulation scenarios in predeployment tool-permission and retrieval testing; a credible check must inject a fake system instruction in a retrieved page and confirm data egress is denied.
For agent cancellation and rollback, it can be sensible to cancel the reservation through its API and reconcile stored workflow state if a customer withdraws consent midway through an agent flow that already performed a reversible reservation. Without the actual change to block the attempted exfiltration and isolate the injected source, the answer lacks the correct operational scope even if the alternate action is technically valid elsewhere.
The action to test tool boundaries against untrusted instructions and independent policy guards can contribute to adversarial instruction testing as a ongoing preventive rule step; here, a retrieved web page tells an agent to ignore its instructions and leak the current customer record. The more directly responsive choice is to block the attempted exfiltration and isolate the injected source.
The action to inject a fake system instruction in a retrieved page and confirm data egress is denied can contribute to adversarial instruction testing as a targeted acceptance check step; here, a retrieved web page tells an agent to ignore its instructions and leak the current customer record. The more directly responsive choice is to block the attempted exfiltration and isolate the injected source.
The action to include prompt-manipulation scenarios in predeployment tool-permission and retrieval testing can contribute to adversarial instruction testing as a implementation step; here, a retrieved web page tells an agent to ignore its instructions and leak the current customer record. The more directly responsive choice is to block the attempted exfiltration and isolate the injected source.
Question 494
A Copilot Studio architect at an insurance provider wants the following issue to be prevented in later deployments: A retrieved web page tells an agent to ignore its instructions and leak the current customer record. Which enduring control or engineering guardrail BEST prevents recurrence?
Explanation
Correct Answer
Standardizing the boundary to test tool boundaries against untrusted instructions and independent policy guards is preventive because the same adversarial instruction testing failure could otherwise return: a retrieved web page tells an agent to ignore its instructions and leak the current customer record. For this insurance provider, the service should also include prompt-manipulation scenarios in predeployment tool-permission and retrieval testing, and tests should inject a fake system instruction in a retrieved page and confirm data egress is denied so that the ongoing requirement is measurable.
Incorrect Answers
The action to require test coverage across users intents languages and failure paths addresses agent evaluation set representativeness rather than the decisive adversarial instruction testing condition. The insurance provider should not infer that fixing one adjacent business-system dependency validates its unrelated workflow; the requested action is to test tool boundaries against untrusted instructions and independent policy guards.
The action to block the attempted exfiltration and isolate the injected source can contribute to adversarial instruction testing as a corrective remediation step; here, a retrieved web page tells an agent to ignore its instructions and leak the current customer record. The more directly responsive choice is to test tool boundaries against untrusted instructions and independent policy guards.
The assumption to treat malicious webpage content as a trusted enterprise system command is the exact failure pattern the team needs to avoid. The relevant change is to include prompt-manipulation scenarios in predeployment tool-permission and retrieval testing; a credible check must inject a fake system instruction in a retrieved page and confirm data egress is denied.
Question 495
During an agent reliability assessment, a Power Platform administrator must accept a change for adversarial instruction testing. The issue to be addressed is: A retrieved web page tells an agent to ignore its instructions and leak the current customer record. Which concrete verification action demonstrates that the relevant behavior works?
Explanation
Correct Answer
The targeted acceptance test is to inject a fake system instruction in a retrieved page and confirm data egress is denied. The design should first include prompt-manipulation scenarios in predeployment tool-permission and retrieval testing. A robust software publisher deployment also needs to test tool boundaries against untrusted instructions and independent policy guards; passing this check provides relevant evidence but does not certify every possible input or security risk.
Incorrect Answers
The action to test tool boundaries against untrusted instructions and independent policy guards can contribute to adversarial instruction testing as a ongoing preventive rule step; here, a retrieved web page tells an agent to ignore its instructions and leak the current customer record. The more directly responsive choice is to inject a fake system instruction in a retrieved page and confirm data egress is denied.
The assumption to treat malicious webpage content as a trusted enterprise system command is the exact failure pattern the team needs to avoid. The relevant change is to include prompt-manipulation scenarios in predeployment tool-permission and retrieval testing; a credible check must inject a fake system instruction in a retrieved page and confirm data egress is denied.
The action to include prompt-manipulation scenarios in predeployment tool-permission and retrieval testing can contribute to adversarial instruction testing as a implementation step; here, a retrieved web page tells an agent to ignore its instructions and leak the current customer record. The more directly responsive choice is to inject a fake system instruction in a retrieved page and confirm data egress is denied.
Choosing to run high-risk and routine queries and inspect selected models and final quality would direct effort toward model router evaluation, whose characteristic issue is that a routing layer sends complex regulated advice to the cheapest small model to save cost. The relevant corrective or preventive action here remains to inject a fake system instruction in a retrieved page and confirm data egress is denied, with its own validation criteria.
Question 496
A model evaluation specialist at a research institute is conducting an incident response exercise concerning agent cancellation and rollback. The observed problem is: A customer withdraws consent midway through an agent flow that already performed a reversible reservation. Which TWO actions together represent the needed implementation AND a durable preventive safeguard? (Select TWO answers.)
Explanation
Correct Answers
Choosing to test cancellation compensation and state reconciliation at each side-effecting step directly targets the described agent cancellation and rollback gap: a customer withdraws consent midway through an agent flow that already performed a reversible reservation. As a lasting boundary, they also need to specify compensation rules and cutoff points for agent-led transactions, because an implementation by itself does not prove safe operation.
Standardizing the boundary to specify compensation rules and cutoff points for agent-led transactions is preventive because the same agent cancellation and rollback failure could otherwise return: a customer withdraws consent midway through an agent flow that already performed a reversible reservation. For this research institute, the service should also test cancellation compensation and state reconciliation at each side-effecting step, and tests should cancel after reservation creation and confirm no orphaned booking remains so that the ongoing requirement is measurable.
Incorrect Answers
The action to cancel after reservation creation and confirm no orphaned booking remains can contribute to agent cancellation and rollback as a targeted acceptance check step; here, a customer withdraws consent midway through an agent flow that already performed a reversible reservation. The more directly responsive choice is to test cancellation compensation and state reconciliation at each side-effecting step.
The option to measure claim-to-citation faithfulness and flag unsupported assertions for review would normally serve a different rag citation faithfulness requirement, where an enterprise assistant cites a relevant document but states a policy exception the document does not contain. The alternative may be useful in its own context, but it does not directly supply test cancellation compensation and state reconciliation at each side-effecting step for the stated decision.
The assumption to assume closing the chat window automatically reverses external transactions is the exact failure pattern the team needs to avoid. The relevant change is to test cancellation compensation and state reconciliation at each side-effecting step; a credible check must cancel after reservation creation and confirm no orphaned booking remains.
Question 497
After a business process modernization at a hospitality operator, the team observes this failure: A customer withdraws consent midway through an agent flow that already performed a reversible reservation. Which flawed assumption or implementation choice would MOST DIRECTLY explain why the intended outcome was not achieved?
Explanation
Correct Answer
The faulty assumption is to assume closing the chat window automatically reverses external transactions. A functioning alternative would be to test cancellation compensation and state reconciliation at each side-effecting step, and the relevant validation would be to cancel after reservation creation and confirm no orphaned booking remains.
Incorrect Answers
The option to run tests solely under an administrator identity with universal access would normally serve a different agent tool authorization negative test requirement, where a role-limited agent can still call a tool action that updates executive compensation data. The alternative may be useful in its own context, but it does not directly supply assume closing the chat window automatically reverses external transactions for the stated decision.
The alternative to cancel after reservation creation and confirm no orphaned booking remains is a targeted acceptance check step rather than the faulty assumption being diagnosed. The failure in agent cancellation and rollback arises from assume closing the chat window automatically reverses external transactions; the concrete symptom is that a customer withdraws consent midway through an agent flow that already performed a reversible reservation.
The alternative to test cancellation compensation and state reconciliation at each side-effecting step is a implementation step rather than the faulty assumption being diagnosed. The failure in agent cancellation and rollback arises from assume closing the chat window automatically reverses external transactions; the concrete symptom is that a customer withdraws consent midway through an agent flow that already performed a reversible reservation.
Question 498
During a security readiness checkpoint at a municipal services department, an AI solution team must address agent cancellation and rollback. The requirement is: A customer withdraws consent midway through an agent flow that already performed a reversible reservation. Which THREE actions together implement the needed solution, its preventive guardrail, and a targeted acceptance test? (Select THREE answers.)
Explanation
Correct Answers
Choosing to test cancellation compensation and state reconciliation at each side-effecting step directly targets the described agent cancellation and rollback gap: a customer withdraws consent midway through an agent flow that already performed a reversible reservation. As a lasting boundary, they also need to specify compensation rules and cutoff points for agent-led transactions, because an implementation by itself does not prove safe operation.
Standardizing the boundary to specify compensation rules and cutoff points for agent-led transactions is preventive because the same agent cancellation and rollback failure could otherwise return: a customer withdraws consent midway through an agent flow that already performed a reversible reservation. For this municipal services department, the service should also test cancellation compensation and state reconciliation at each side-effecting step, and tests should cancel after reservation creation and confirm no orphaned booking remains so that the ongoing requirement is measurable.
To verify this change, the municipal services department should cancel after reservation creation and confirm no orphaned booking remains. Operators still need the implementation to test cancellation compensation and state reconciliation at each side-effecting step and the preventive requirement to specify compensation rules and cutoff points for agent-led transactions.
Incorrect Answers
The action to cancel the reservation through its API and reconcile stored workflow state can contribute to agent cancellation and rollback as a corrective remediation step; here, a customer withdraws consent midway through an agent flow that already performed a reversible reservation. The more directly responsive choice is to test cancellation compensation and state reconciliation at each side-effecting step.
For agent evaluation set representativeness, it can be sensible to require test coverage across users intents languages and failure paths if an agent is evaluated only on easy happy-path prompts despite frequent ambiguous production requests. Without the actual change to test cancellation compensation and state reconciliation at each side-effecting step, the answer lacks the correct operational scope even if the alternate action is technically valid elsewhere.
The assumption to assume closing the chat window automatically reverses external transactions is the exact failure pattern the team needs to avoid. The relevant change is to test cancellation compensation and state reconciliation at each side-effecting step; a credible check must cancel after reservation creation and confirm no orphaned booking remains.
Question 499
A Dynamics 365 consultant at a global distributor wants the following issue to be prevented in later deployments: A customer withdraws consent midway through an agent flow that already performed a reversible reservation. Which enduring control or engineering guardrail BEST prevents recurrence?
Explanation
Correct Answer
The sustained prevention measure is to specify compensation rules and cutoff points for agent-led transactions. The implementation can still use the separate action to test cancellation compensation and state reconciliation at each side-effecting step, but a guardrail constrains future changes. The global distributor can measure whether the rule works by arranging to cancel after reservation creation and confirm no orphaned booking remains.
Incorrect Answers
The action to cancel the reservation through its API and reconcile stored workflow state can contribute to agent cancellation and rollback as a corrective remediation step; here, a customer withdraws consent midway through an agent flow that already performed a reversible reservation. The more directly responsive choice is to specify compensation rules and cutoff points for agent-led transactions.
The assumption to assume closing the chat window automatically reverses external transactions is the exact failure pattern the team needs to avoid. The relevant change is to test cancellation compensation and state reconciliation at each side-effecting step; a credible check must cancel after reservation creation and confirm no orphaned booking remains.
The action to require automated access-denied checks at the actual tool service addresses agent tool authorization negative test rather than the decisive agent cancellation and rollback condition. The global distributor should not infer that fixing one adjacent business-system dependency validates its unrelated workflow; the requested action is to specify compensation rules and cutoff points for agent-led transactions.
Question 500
During a business process modernization, a customer service architect must accept a change for agent cancellation and rollback. The issue to be addressed is: A customer withdraws consent midway through an agent flow that already performed a reversible reservation. Which concrete verification action demonstrates that the relevant behavior works?
Explanation
Correct Answer
To verify this change, the digital payments provider should cancel after reservation creation and confirm no orphaned booking remains. Operators still need the implementation to test cancellation compensation and state reconciliation at each side-effecting step and the preventive requirement to specify compensation rules and cutoff points for agent-led transactions.
Incorrect Answers
For fairness threshold validation, it can be sensible to compare acceptance rates and error labels across relevant supplier groups if an application scores supplier proposals but a vulnerable group has disproportionate false rejections. Without the actual change to cancel after reservation creation and confirm no orphaned booking remains, the answer lacks the correct operational scope even if the alternate action is technically valid elsewhere.
The assumption to assume closing the chat window automatically reverses external transactions is the exact failure pattern the team needs to avoid. The relevant change is to test cancellation compensation and state reconciliation at each side-effecting step; a credible check must cancel after reservation creation and confirm no orphaned booking remains.
The action to test cancellation compensation and state reconciliation at each side-effecting step can contribute to agent cancellation and rollback as a implementation step; here, a customer withdraws consent midway through an agent flow that already performed a reversible reservation. The more directly responsive choice is to cancel after reservation creation and confirm no orphaned booking remains.
Agentic AI Business Solutions Architect (AB-100) Practice Tests
- Microsoft AB-100 Practice Test 48 - 10 Questions (Last Reviewed 05-Oct-2026)
- Microsoft AB-100 Practice Test 49 - 10 Questions (Last Reviewed 04-Oct-2026)
- Microsoft AB-100 Practice Test 50 - 10 Questions (Last Reviewed 03-Oct-2026)
- Microsoft AB-100 Practice Test 51 - 10 Questions (Last Reviewed 02-Oct-2026)
- Microsoft AB-100 Practice Test 52 - 10 Questions (Last Reviewed 01-Oct-2026)