IAPP AIGP Practice Test 108
Question 1070
A live recruitment assistant sends rejection notices automatically while reviewers can only see decisions after dispatch. Postdeployment review has identified the same issue after a material service change. What action should the accountable team prioritize to govern this use responsibly?
- Require qualified review and an effective stop/override before adverse employment actions.
- Create an external communication plan covering AI involvement, limitations and human escalation.
- Require a new use-case review and prevent unsupported profiling from accessing summary data.
- Contain the tool, investigate exposure and coordinate security, legal, privacy and customer communications.
Explanation
Correct Answer
A:Require qualified review and an effective stop/override before adverse employment actions. Correct. It directly reduces or governs the specific documented risk. Human oversight at deployment must be timely, usable and proportionate to consequences.
Incorrect Answers
B:Create an external communication plan covering AI involvement, limitations and human escalation. Not selected. That action treats another control objective, not the specific gap described. It belongs with External AI disclosure: Transparency requires practical communication and redress suited to downstream users. The present case instead concerns Production human oversight.
C:Require a new use-case review and prevent unsupported profiling from accessing summary data. Not selected. That action treats another control objective, not the specific gap described. It belongs with Secondary use and function creep: A low-risk starting use can become high-risk when repurposed for consequential decisions. The present case instead concerns Production human oversight.
D:Contain the tool, investigate exposure and coordinate security, legal, privacy and customer communications. Not selected. That action treats another control objective, not the specific gap described. It belongs with Incident response and notification: An AI incident response combines technical containment with organizational and legal obligations. The present case instead concerns Production human oversight.
Question 1071
A live recruitment assistant sends rejection notices automatically while reviewers can only see decisions after dispatch. Postdeployment review has identified the same issue after a material service change. What result would provide the strongest case-specific assurance?
- Test a contested case and confirm the reviewer can pause, explain and reverse an action.
- Validate that downstream systems switch safely and no shadow users keep relying on the old model.
- Measure misclassification and complaints by language and region under real operations.
- Test affected and unaffected jurisdictions for correct blocking and no easy bypass.
Explanation
Correct Answer
A:Test a contested case and confirm the reviewer can pause, explain and reverse an action. Correct. It yields measurable evidence that the case-specific safeguard works. Human oversight at deployment must be timely, usable and proportionate to consequences.
Incorrect Answers
B:Validate that downstream systems switch safely and no shadow users keep relying on the old model. Not selected. That test would not demonstrate this particular risk has been addressed. It belongs with Model retirement and replacement: Decommissioning is a lifecycle control that prevents indefinite reliance on a failing system. The present case instead concerns Production human oversight.
C:Measure misclassification and complaints by language and region under real operations. Not selected. That test would not demonstrate this particular risk has been addressed. It belongs with Distribution shift detection: A geographic or population expansion can invalidate the original model's evidence. The present case instead concerns Production human oversight.
D:Test affected and unaffected jurisdictions for correct blocking and no easy bypass. Not selected. That test would not demonstrate this particular risk has been addressed. It belongs with Jurisdictional localization: Responsible deployment may require localized disabling rather than a global binary shutdown. The present case instead concerns Production human oversight.
Question 1072
A live recruitment assistant sends rejection notices automatically while reviewers can only see decisions after dispatch. Postdeployment review has identified the same issue after a material service change. Which TWO actions form the appropriate investigation-and-control pair for this case?
- Inspect the production action point, reviewer authority, workload and reversibility of adverse outcomes.
- Require qualified review and an effective stop/override before adverse employment actions.
- Assess how user behavior, automation bias and inadequate instructions create operational harms.
- Gate new-region deployment on representative testing and an updated impact assessment.
Explanation
Correct Answers
A:Inspect the production action point, reviewer authority, workload and reversibility of adverse outcomes. Correct. As the first, evidence-gathering half of the requested pair, this establishes the specific risk boundary. Human oversight at deployment must be timely, usable and proportionate to consequences.
B:Require qualified review and an effective stop/override before adverse employment actions. Correct. As the second, implementation half of the requested pair, this directly applies the needed safeguard. Human oversight at deployment must be timely, usable and proportionate to consequences.
Incorrect Answers
C:Assess how user behavior, automation bias and inadequate instructions create operational harms. Not selected. It addresses a different governance mechanism, not the required assessment-and-control pair. It belongs with User training and instructions: Even a well-tested model can be unsafe when deployers or users misunderstand its limitations. The present case instead concerns Production human oversight.
D:Gate new-region deployment on representative testing and an updated impact assessment. Not selected. It addresses a different governance mechanism, not the required assessment-and-control pair. It belongs with Distribution shift detection: A geographic or population expansion can invalidate the original model's evidence. The present case instead concerns Production human oversight.
Question 1073
Frontline staff rely on an AI decision aid without understanding confidence limits, inappropriate uses or how to report errors. The operational owner must decide whether the current deployment can safely proceed. Before selecting a remedy, what should the accountable reviewer investigate first?
- Assess how user behavior, automation bias and inadequate instructions create operational harms.
- Find privileged credentials, execution queues, decision gates and current disablement coverage.
- Check changes in connected tools, attack surface, permissions and incident trends since the last test.
- Identify who needs accurate AI-use disclosures, contest routes and contact instructions.
Explanation
Correct Answer
A:Assess how user behavior, automation bias and inadequate instructions create operational harms. Correct. It establishes the relevant factual risk boundary before implementation. Even a well-tested model can be unsafe when deployers or users misunderstand its limitations.
Incorrect Answers
B:Find privileged credentials, execution queues, decision gates and current disablement coverage. Not selected. That check primarily diagnoses a different governance issue. It belongs with Deactivation and kill switch: High-autonomy AI needs effective interruption and containment, not merely an on-screen warning. The present case instead concerns User training and instructions.
C:Check changes in connected tools, attack surface, permissions and incident trends since the last test. Not selected. That check primarily diagnoses a different governance issue. It belongs with Periodic red teaming in operation: Changes in deployment context can reintroduce adversarial risks after launch testing. The present case instead concerns User training and instructions.
D:Identify who needs accurate AI-use disclosures, contest routes and contact instructions. Not selected. That check primarily diagnoses a different governance issue. It belongs with External AI disclosure: Transparency requires practical communication and redress suited to downstream users. The present case instead concerns User training and instructions.
Question 1074
Frontline staff rely on an AI decision aid without understanding confidence limits, inappropriate uses or how to report errors. The operational owner must decide whether the current deployment can safely proceed. What action should the accountable team prioritize to govern this use responsibly?
- Require qualified review and an effective stop/override before adverse employment actions.
- Assign measurable remediation, owner and retest deadlines with independent verification.
- Gate new-region deployment on representative testing and an updated impact assessment.
- Train users on validated operating domain, uncertainty, escalation and prohibited shortcuts.
Explanation
Correct Answer
D:Train users on validated operating domain, uncertainty, escalation and prohibited shortcuts. Correct. It directly reduces or governs the specific documented risk. Even a well-tested model can be unsafe when deployers or users misunderstand its limitations.
Incorrect Answers
A:Require qualified review and an effective stop/override before adverse employment actions. Not selected. That action treats another control objective, not the specific gap described. It belongs with Production human oversight: Human oversight at deployment must be timely, usable and proportionate to consequences. The present case instead concerns User training and instructions.
B:Assign measurable remediation, owner and retest deadlines with independent verification. Not selected. That action treats another control objective, not the specific gap described. It belongs with Continuous improvement and accountability: Continual improvement depends on verified risk reduction, not the number of revised policy documents. The present case instead concerns User training and instructions.
C:Gate new-region deployment on representative testing and an updated impact assessment. Not selected. That action treats another control objective, not the specific gap described. It belongs with Distribution shift detection: A geographic or population expansion can invalidate the original model's evidence. The present case instead concerns User training and instructions.
Question 1075
Frontline staff rely on an AI decision aid without understanding confidence limits, inappropriate uses or how to report errors. The operational owner must decide whether the current deployment can safely proceed. What result would provide the strongest case-specific assurance?
- Test authorization boundaries and verify disciplinary users cannot access prohibited scoring features.
- Test a contested case and confirm the reviewer can pause, explain and reverse an action.
- Use scenario-based proficiency tests and inspect actual help-seeking and override behavior.
- Simulate a compromise and confirm tool calls stop within the required time and leave auditable records.
Explanation
Correct Answer
C:Use scenario-based proficiency tests and inspect actual help-seeking and override behavior. Correct. It yields measurable evidence that the case-specific safeguard works. Even a well-tested model can be unsafe when deployers or users misunderstand its limitations.
Incorrect Answers
A:Test authorization boundaries and verify disciplinary users cannot access prohibited scoring features. Not selected. That test would not demonstrate this particular risk has been addressed. It belongs with Secondary use and function creep: A low-risk starting use can become high-risk when repurposed for consequential decisions. The present case instead concerns User training and instructions.
B:Test a contested case and confirm the reviewer can pause, explain and reverse an action. Not selected. That test would not demonstrate this particular risk has been addressed. It belongs with Production human oversight: Human oversight at deployment must be timely, usable and proportionate to consequences. The present case instead concerns User training and instructions.
D:Simulate a compromise and confirm tool calls stop within the required time and leave auditable records. Not selected. That test would not demonstrate this particular risk has been addressed. It belongs with Deactivation and kill switch: High-autonomy AI needs effective interruption and containment, not merely an on-screen warning. The present case instead concerns User training and instructions.
Question 1076
Frontline staff rely on an AI decision aid without understanding confidence limits, inappropriate uses or how to report errors. The operational owner must decide whether the current deployment can safely proceed. What should the owner do to prevent further harm from this specific failure?
- Temporarily restrict untrained users from consequential AI functions.
- Deactivate the restricted feature in the relevant jurisdiction until compliance is established.
- Retire the unsafe model and invoke the authorized fallback while preserving decision records.
- Disable automatic rejection and use manual review pending effective controls.
Explanation
Correct Answer
A:Temporarily restrict untrained users from consequential AI functions. Correct. It limits further reliance or exposure while the risk remains unresolved. Even a well-tested model can be unsafe when deployers or users misunderstand its limitations.
Incorrect Answers
B:Deactivate the restricted feature in the relevant jurisdiction until compliance is established. Not selected. That response contains a different failure mode and leaves the stated risk unresolved. It belongs with Jurisdictional localization: Responsible deployment may require localized disabling rather than a global binary shutdown. The present case instead concerns User training and instructions.
C:Retire the unsafe model and invoke the authorized fallback while preserving decision records. Not selected. That response contains a different failure mode and leaves the stated risk unresolved. It belongs with Model retirement and replacement: Decommissioning is a lifecycle control that prevents indefinite reliance on a failing system. The present case instead concerns User training and instructions.
D:Disable automatic rejection and use manual review pending effective controls. Not selected. That response contains a different failure mode and leaves the stated risk unresolved. It belongs with Production human oversight: Human oversight at deployment must be timely, usable and proportionate to consequences. The present case instead concerns User training and instructions.
Question 1077
Frontline staff rely on an AI decision aid without understanding confidence limits, inappropriate uses or how to report errors. A new business unit wants to scale the system without a further governance review. What should the AI governance lead examine before changing the system?
- Compare the new use with approved purpose, law, stakeholder expectations and plausible downstream harm.
- Assess how user behavior, automation bias and inadequate instructions create operational harms.
- Analyze distribution differences, language coverage and subgroup calibration before expanding reach.
- Review model changes, field incidents, monitoring results and the post-market documentation baseline.
Explanation
Correct Answer
B:Assess how user behavior, automation bias and inadequate instructions create operational harms. Correct. It establishes the relevant factual risk boundary before implementation. Even a well-tested model can be unsafe when deployers or users misunderstand its limitations.
Incorrect Answers
A:Compare the new use with approved purpose, law, stakeholder expectations and plausible downstream harm. Not selected. That check primarily diagnoses a different governance issue. It belongs with Secondary use and function creep: A low-risk starting use can become high-risk when repurposed for consequential decisions. The present case instead concerns User training and instructions.
C:Analyze distribution differences, language coverage and subgroup calibration before expanding reach. Not selected. That check primarily diagnoses a different governance issue. It belongs with Distribution shift detection: A geographic or population expansion can invalidate the original model's evidence. The present case instead concerns User training and instructions.
D:Review model changes, field incidents, monitoring results and the post-market documentation baseline. Not selected. That check primarily diagnoses a different governance issue. It belongs with Postmarket documentation: AI lifecycle governance requires versioned post-market evidence, not only prelaunch paperwork. The present case instead concerns User training and instructions.
Question 1078
Frontline staff rely on an AI decision aid without understanding confidence limits, inappropriate uses or how to report errors. A new business unit wants to scale the system without a further governance review. What is the most directly applicable governance control for this situation?
- Create an external communication plan covering AI involvement, limitations and human escalation.
- Contain the tool, investigate exposure and coordinate security, legal, privacy and customer communications.
- Train users on validated operating domain, uncertainty, escalation and prohibited shortcuts.
- Require a new use-case review and prevent unsupported profiling from accessing summary data.
Explanation
Correct Answer
C:Train users on validated operating domain, uncertainty, escalation and prohibited shortcuts. Correct. It directly reduces or governs the specific documented risk. Even a well-tested model can be unsafe when deployers or users misunderstand its limitations.
Incorrect Answers
A:Create an external communication plan covering AI involvement, limitations and human escalation. Not selected. That action treats another control objective, not the specific gap described. It belongs with External AI disclosure: Transparency requires practical communication and redress suited to downstream users. The present case instead concerns User training and instructions.
B:Contain the tool, investigate exposure and coordinate security, legal, privacy and customer communications. Not selected. That action treats another control objective, not the specific gap described. It belongs with Incident response and notification: An AI incident response combines technical containment with organizational and legal obligations. The present case instead concerns User training and instructions.
D:Require a new use-case review and prevent unsupported profiling from accessing summary data. Not selected. That action treats another control objective, not the specific gap described. It belongs with Secondary use and function creep: A low-risk starting use can become high-risk when repurposed for consequential decisions. The present case instead concerns User training and instructions.
Question 1079
Frontline staff rely on an AI decision aid without understanding confidence limits, inappropriate uses or how to report errors. A new business unit wants to scale the system without a further governance review. What should an independent reviewer test before signing off on this specific risk?
- Reproduce the exploit in a safe environment and verify least-privilege remediation.
- Use scenario-based proficiency tests and inspect actual help-seeking and override behavior.
- Use follow-up audits to confirm the control works on live workflows and recurring findings decline.
- Test a contested case and confirm the reviewer can pause, explain and reverse an action.
Explanation
Correct Answer
B:Use scenario-based proficiency tests and inspect actual help-seeking and override behavior. Correct. It yields measurable evidence that the case-specific safeguard works. Even a well-tested model can be unsafe when deployers or users misunderstand its limitations.
Incorrect Answers
A:Reproduce the exploit in a safe environment and verify least-privilege remediation. Not selected. That test would not demonstrate this particular risk has been addressed. It belongs with Incident response and notification: An AI incident response combines technical containment with organizational and legal obligations. The present case instead concerns User training and instructions.
C:Use follow-up audits to confirm the control works on live workflows and recurring findings decline. Not selected. That test would not demonstrate this particular risk has been addressed. It belongs with Continuous improvement and accountability: Continual improvement depends on verified risk reduction, not the number of revised policy documents. The present case instead concerns User training and instructions.
D:Test a contested case and confirm the reviewer can pause, explain and reverse an action. Not selected. That test would not demonstrate this particular risk has been addressed. It belongs with Production human oversight: Human oversight at deployment must be timely, usable and proportionate to consequences. The present case instead concerns User training and instructions.
Question 1080
Frontline staff rely on an AI decision aid without understanding confidence limits, inappropriate uses or how to report errors. A new business unit wants to scale the system without a further governance review. What is the most relevant fallback when the promised safeguard cannot be demonstrated?
- Disable automatic rejection and use manual review pending effective controls.
- Disable the vulnerable connector and revoke excessive permissions pending remediation.
- Reopen falsely closed findings and restrict affected high-risk operations when controls remain ineffective.
- Temporarily restrict untrained users from consequential AI functions.
Explanation
Correct Answer
D:Temporarily restrict untrained users from consequential AI functions. Correct. It limits further reliance or exposure while the risk remains unresolved. Even a well-tested model can be unsafe when deployers or users misunderstand its limitations.
Incorrect Answers
A:Disable automatic rejection and use manual review pending effective controls. Not selected. That response contains a different failure mode and leaves the stated risk unresolved. It belongs with Production human oversight: Human oversight at deployment must be timely, usable and proportionate to consequences. The present case instead concerns User training and instructions.
B:Disable the vulnerable connector and revoke excessive permissions pending remediation. Not selected. That response contains a different failure mode and leaves the stated risk unresolved. It belongs with Periodic red teaming in operation: Changes in deployment context can reintroduce adversarial risks after launch testing. The present case instead concerns User training and instructions.
C:Reopen falsely closed findings and restrict affected high-risk operations when controls remain ineffective. Not selected. That response contains a different failure mode and leaves the stated risk unresolved. It belongs with Continuous improvement and accountability: Continual improvement depends on verified risk reduction, not the number of revised policy documents. The present case instead concerns User training and instructions.