IAPP AIGP Practice Test 114
Question 1130
An internal workplace assistant built for meeting summaries is proposed for disciplinary profiling without employee consultation. Postdeployment review has identified the same issue after a material service change. What action should the accountable team prioritize to govern this use responsibly?
- Implement jurisdiction-aware feature gating with documented legal review and auditable location controls.
- Require qualified review and an effective stop/override before adverse employment actions.
- Require a new use-case review and prevent unsupported profiling from accessing summary data.
- Update model version records, operating instructions, post-market plan and relevant compliance evidence.
Explanation
Correct Answer
C:Require a new use-case review and prevent unsupported profiling from accessing summary data. Correct. It directly reduces or governs the specific documented risk. A low-risk starting use can become high-risk when repurposed for consequential decisions.
Incorrect Answers
A:Implement jurisdiction-aware feature gating with documented legal review and auditable location controls. Not selected. That action treats another control objective, not the specific gap described. It belongs with Jurisdictional localization: Responsible deployment may require localized disabling rather than a global binary shutdown. The present case instead concerns Secondary use and function creep.
B: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 Secondary use and function creep.
D:Update model version records, operating instructions, post-market plan and relevant compliance evidence. Not selected. That action treats another control objective, not the specific gap described. It belongs with Postmarket documentation: AI lifecycle governance requires versioned post-market evidence, not only prelaunch paperwork. The present case instead concerns Secondary use and function creep.
Question 1131
An internal workplace assistant built for meeting summaries is proposed for disciplinary profiling without employee consultation. Postdeployment review has identified the same issue after a material service change. What result would provide the strongest case-specific assurance?
- Measure misclassification and complaints by language and region under real operations.
- Sample real incidents and verify traceability to actions, owners and updated disclosures.
- Attempt controlled prompt injection through retrieved content and confirm confidential data is not exposed.
- Test authorization boundaries and verify disciplinary users cannot access prohibited scoring features.
Explanation
Correct Answer
D:Test authorization boundaries and verify disciplinary users cannot access prohibited scoring features. Correct. It yields measurable evidence that the case-specific safeguard works. A low-risk starting use can become high-risk when repurposed for consequential decisions.
Incorrect Answers
A: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 Secondary use and function creep.
B:Sample real incidents and verify traceability to actions, owners and updated disclosures. Not selected. That test would not demonstrate this particular risk has been addressed. It belongs with Postmarket documentation: AI lifecycle governance requires versioned post-market evidence, not only prelaunch paperwork. The present case instead concerns Secondary use and function creep.
C:Attempt controlled prompt injection through retrieved content and confirm confidential data is not exposed. Not selected. That test would not demonstrate this particular risk has been addressed. It belongs with Periodic red teaming in operation: Changes in deployment context can reintroduce adversarial risks after launch testing. The present case instead concerns Secondary use and function creep.
Question 1132
An internal workplace assistant built for meeting summaries is proposed for disciplinary profiling without employee consultation. 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?
- Identify unacceptable risk trends, downstream dependencies, replacement criteria and the status of a safe fallback.
- Compare the new use with approved purpose, law, stakeholder expectations and plausible downstream harm.
- Gate new-region deployment on representative testing and an updated impact assessment.
- Require a new use-case review and prevent unsupported profiling from accessing summary data.
Explanation
Correct Answers
B:Compare the new use with approved purpose, law, stakeholder expectations and plausible downstream harm. Correct. As the first, evidence-gathering half of the requested pair, this establishes the specific risk boundary. A low-risk starting use can become high-risk when repurposed for consequential decisions.
D:Require a new use-case review and prevent unsupported profiling from accessing summary data. Correct. As the second, implementation half of the requested pair, this directly applies the needed safeguard. A low-risk starting use can become high-risk when repurposed for consequential decisions.
Incorrect Answers
A:Identify unacceptable risk trends, downstream dependencies, replacement criteria and the status of a safe fallback. Not selected. It addresses a different governance mechanism, not the required assessment-and-control pair. 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 Secondary use and function creep.
C: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 Secondary use and function creep.
Question 1133
An AI supplier replaces a clinical model after deployment but does not update the risk register or operating instructions. The operational owner must decide whether the current deployment can safely proceed. Before selecting a remedy, what should the accountable reviewer investigate first?
- Identify who needs accurate AI-use disclosures, contest routes and contact instructions.
- Review model changes, field incidents, monitoring results and the post-market documentation baseline.
- Check changes in connected tools, attack surface, permissions and incident trends since the last test.
- Identify unacceptable risk trends, downstream dependencies, replacement criteria and the status of a safe fallback.
Explanation
Correct Answer
B:Review model changes, field incidents, monitoring results and the post-market documentation baseline. Correct. It establishes the relevant factual risk boundary before implementation. AI lifecycle governance requires versioned post-market evidence, not only prelaunch paperwork.
Incorrect Answers
A: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 Postmarket documentation.
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 Postmarket documentation.
D:Identify unacceptable risk trends, downstream dependencies, replacement criteria and the status of a safe fallback. Not selected. That check primarily diagnoses a different governance issue. 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 Postmarket documentation.
Question 1134
An AI supplier replaces a clinical model after deployment but does not update the risk register or operating instructions. 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.
- Update model version records, operating instructions, post-market plan and relevant compliance evidence.
- Monitor balanced outcome and harm indicators with thresholds and accountable escalation.
- Schedule risk-triggered adversarial testing, threat modeling and security audits after material changes.
Explanation
Correct Answer
B:Update model version records, operating instructions, post-market plan and relevant compliance evidence. Correct. It directly reduces or governs the specific documented risk. AI lifecycle governance requires versioned post-market evidence, not only prelaunch paperwork.
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 Postmarket documentation.
C:Monitor balanced outcome and harm indicators with thresholds and accountable escalation. Not selected. That action treats another control objective, not the specific gap described. It belongs with Postdeployment outcome monitoring: Postdeployment success includes real-world harms and stakeholder outcomes, not just engagement. The present case instead concerns Postmarket documentation.
D:Schedule risk-triggered adversarial testing, threat modeling and security audits after material changes. Not selected. That action treats another control objective, not the specific gap described. It belongs with Periodic red teaming in operation: Changes in deployment context can reintroduce adversarial risks after launch testing. The present case instead concerns Postmarket documentation.
Question 1135
An AI supplier replaces a clinical model after deployment but does not update the risk register or operating instructions. The operational owner must decide whether the current deployment can safely proceed. What result would provide the strongest case-specific assurance?
- Use scenario-based proficiency tests and inspect actual help-seeking and override behavior.
- Sample real incidents and verify traceability to actions, owners and updated disclosures.
- Validate that downstream systems switch safely and no shadow users keep relying on the old model.
- Simulate a compromise and confirm tool calls stop within the required time and leave auditable records.
Explanation
Correct Answer
B:Sample real incidents and verify traceability to actions, owners and updated disclosures. Correct. It yields measurable evidence that the case-specific safeguard works. AI lifecycle governance requires versioned post-market evidence, not only prelaunch paperwork.
Incorrect Answers
A:Use scenario-based proficiency tests and inspect actual help-seeking and override behavior. Not selected. That test would not demonstrate this particular risk has been addressed. 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 Postmarket documentation.
C: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 Postmarket documentation.
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 Postmarket documentation.
Question 1136
An AI supplier replaces a clinical model after deployment but does not update the risk register or operating instructions. 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?
- Correct misleading communications and offer a human review path for affected decisions.
- Retire the unsafe model and invoke the authorized fallback while preserving decision records.
- Freeze further unreviewed model updates until postmarket records are reconciled.
- Reopen falsely closed findings and restrict affected high-risk operations when controls remain ineffective.
Explanation
Correct Answer
C:Freeze further unreviewed model updates until postmarket records are reconciled. Correct. It limits further reliance or exposure while the risk remains unresolved. AI lifecycle governance requires versioned post-market evidence, not only prelaunch paperwork.
Incorrect Answers
A:Correct misleading communications and offer a human review path for affected decisions. Not selected. That response contains a different failure mode and leaves the stated risk unresolved. It belongs with External AI disclosure: Transparency requires practical communication and redress suited to downstream users. The present case instead concerns Postmarket documentation.
B: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 Postmarket documentation.
D: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 Postmarket documentation.
Question 1137
An AI supplier replaces a clinical model after deployment but does not update the risk register or operating instructions. 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?
- Determine which law, service feature, user population and data routing are affected by the locality restriction.
- Analyze distribution differences, language coverage and subgroup calibration before expanding reach.
- Check changes in connected tools, attack surface, permissions and incident trends since the last test.
- Review model changes, field incidents, monitoring results and the post-market documentation baseline.
Explanation
Correct Answer
D:Review model changes, field incidents, monitoring results and the post-market documentation baseline. Correct. It establishes the relevant factual risk boundary before implementation. AI lifecycle governance requires versioned post-market evidence, not only prelaunch paperwork.
Incorrect Answers
A:Determine which law, service feature, user population and data routing are affected by the locality restriction. Not selected. That check primarily diagnoses a different governance issue. It belongs with Jurisdictional localization: Responsible deployment may require localized disabling rather than a global binary shutdown. The present case instead concerns Postmarket documentation.
B: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 Postmarket documentation.
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 Postmarket documentation.
Question 1138
An AI supplier replaces a clinical model after deployment but does not update the risk register or operating instructions. 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?
- Adopt an approved retirement plan with transition testing, stakeholder notice and retained evidence.
- Implement jurisdiction-aware feature gating with documented legal review and auditable location controls.
- Update model version records, operating instructions, post-market plan and relevant compliance evidence.
- Create an external communication plan covering AI involvement, limitations and human escalation.
Explanation
Correct Answer
C:Update model version records, operating instructions, post-market plan and relevant compliance evidence. Correct. It directly reduces or governs the specific documented risk. AI lifecycle governance requires versioned post-market evidence, not only prelaunch paperwork.
Incorrect Answers
A:Adopt an approved retirement plan with transition testing, stakeholder notice and retained evidence. Not selected. That action treats another control objective, not the specific gap described. 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 Postmarket documentation.
B:Implement jurisdiction-aware feature gating with documented legal review and auditable location controls. Not selected. That action treats another control objective, not the specific gap described. It belongs with Jurisdictional localization: Responsible deployment may require localized disabling rather than a global binary shutdown. The present case instead concerns Postmarket documentation.
D: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 Postmarket documentation.
Question 1139
An AI supplier replaces a clinical model after deployment but does not update the risk register or operating instructions. 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?
- Test affected customer journeys for understandable notice, contact access and timely response.
- Test authorization boundaries and verify disciplinary users cannot access prohibited scoring features.
- Sample real incidents and verify traceability to actions, owners and updated disclosures.
- Measure misclassification and complaints by language and region under real operations.
Explanation
Correct Answer
C:Sample real incidents and verify traceability to actions, owners and updated disclosures. Correct. It yields measurable evidence that the case-specific safeguard works. AI lifecycle governance requires versioned post-market evidence, not only prelaunch paperwork.
Incorrect Answers
A:Test affected customer journeys for understandable notice, contact access and timely response. Not selected. That test would not demonstrate this particular risk has been addressed. It belongs with External AI disclosure: Transparency requires practical communication and redress suited to downstream users. The present case instead concerns Postmarket documentation.
B: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 Postmarket documentation.
D: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 Postmarket documentation.
Question 1140
An AI supplier replaces a clinical model after deployment but does not update the risk register or operating instructions. 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?
- Freeze further unreviewed model updates until postmarket records are reconciled.
- Reopen falsely closed findings and restrict affected high-risk operations when controls remain ineffective.
- Roll back unsupported languages and provide alternative claims review.
- Reduce or suspend harmful personalization patterns until risks are mitigated.
Explanation
Correct Answer
A:Freeze further unreviewed model updates until postmarket records are reconciled. Correct. It limits further reliance or exposure while the risk remains unresolved. AI lifecycle governance requires versioned post-market evidence, not only prelaunch paperwork.
Incorrect Answers
B: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 Postmarket documentation.
C:Roll back unsupported languages and provide alternative claims review. Not selected. That response contains a different failure mode and leaves the stated risk unresolved. It belongs with Distribution shift detection: A geographic or population expansion can invalidate the original model's evidence. The present case instead concerns Postmarket documentation.
D:Reduce or suspend harmful personalization patterns until risks are mitigated. Not selected. That response contains a different failure mode and leaves the stated risk unresolved. It belongs with Postdeployment outcome monitoring: Postdeployment success includes real-world harms and stakeholder outcomes, not just engagement. The present case instead concerns Postmarket documentation.