IAPP AIGP Practice Test 40
Question 390
A manufacturer embeds computer vision in a robotic safety stop but tests only a limited lighting condition. An independent audit has requested current supporting evidence before expanded use. Which specific implementation best addresses the identified gap?
- Remove unjustified proxy features and establish defensible credit-relevance and adverse-action review controls.
- Require validated fail-safe design, representative safety testing and defect-management controls.
- Require human clearance and vendor-license review for commercial deliverables before release.
- Add job-relevant validation, bias mitigation, documented accommodations and accountable human review.
Explanation
Correct Answer
B:Require validated fail-safe design, representative safety testing and defect-management controls. Correct. It directly reduces or governs the specific documented risk. AI systems inside products can raise ordinary safety and product-liability risks.
Incorrect Answers
A:Remove unjustified proxy features and establish defensible credit-relevance and adverse-action review controls. Not selected. That action treats another control objective, not the specific gap described. It belongs with Credit and lending fairness: AI credit decisions remain subject to lending fairness, transparency and contestability obligations. The present case instead concerns Product safety liability.
C:Require human clearance and vendor-license review for commercial deliverables before release. Not selected. That action treats another control objective, not the specific gap described. It belongs with Generated-output IP: AI output availability does not guarantee ownership, exclusivity or freedom from infringement. The present case instead concerns Product safety liability.
D:Add job-relevant validation, bias mitigation, documented accommodations and accountable human review. Not selected. That action treats another control objective, not the specific gap described. It belongs with Employment discrimination: Equal treatment obligations apply to algorithmic employment decisions and proxy discrimination. The present case instead concerns Product safety liability.
Question 391
A manufacturer embeds computer vision in a robotic safety stop but tests only a limited lighting condition. An independent audit has requested current supporting evidence before expanded use. Which acceptance test most directly verifies the relevant control?
- Sample training records to verify each restricted source has a documented permitted basis.
- Run targeted similarity checks against known protected brands and retain reviewer signoff.
- Determine source licenses, access conditions, copyright exceptions and jurisdiction-specific permissions.
- Test hazardous edge cases and record whether safe-stop performance meets the accepted criterion.
Explanation
Correct Answer
D:Test hazardous edge cases and record whether safe-stop performance meets the accepted criterion. Correct. It yields measurable evidence that the case-specific safeguard works. AI systems inside products can raise ordinary safety and product-liability risks.
Incorrect Answers
A:Sample training records to verify each restricted source has a documented permitted basis. Not selected. That test would not demonstrate this particular risk has been addressed. It belongs with Training-data copyright: Public accessibility does not automatically grant permission to reproduce or train on protected material. The present case instead concerns Product safety liability.
B:Run targeted similarity checks against known protected brands and retain reviewer signoff. Not selected. That test would not demonstrate this particular risk has been addressed. It belongs with Generated-output IP: AI output availability does not guarantee ownership, exclusivity or freedom from infringement. The present case instead concerns Product safety liability.
C:Determine source licenses, access conditions, copyright exceptions and jurisdiction-specific permissions. Not selected. This action relates to Training-data copyright. Public accessibility does not automatically grant permission to reproduce or train on protected material. For the validate decision, it does not resolve the specifically described issue of Product safety liability in A manufacturer embeds computer vision in a robotic safety stop but tests only a limited lighting condition. An independent audit has requested current supporting evidence before expanded use.
Question 392
A manufacturer embeds computer vision in a robotic safety stop but tests only a limited lighting condition. An independent audit has requested current supporting evidence before expanded use. For a defensible decision, choose TWO: one diagnosis and one controlled response.
- Investigate disparate error rates, feature proxies and applicable employment nondiscrimination requirements.
- Identify foreseeable misuse, design defects, failure modes and relevant product safety duties.
- Require validated fail-safe design, representative safety testing and defect-management controls.
- Require human clearance and vendor-license review for commercial deliverables before release.
Explanation
Correct Answers
B:Identify foreseeable misuse, design defects, failure modes and relevant product safety duties. Correct. As the first, evidence-gathering half of the requested pair, this establishes the specific risk boundary. AI systems inside products can raise ordinary safety and product-liability risks.
C:Require validated fail-safe design, representative safety testing and defect-management controls. Correct. As the second, implementation half of the requested pair, this directly applies the needed safeguard. AI systems inside products can raise ordinary safety and product-liability risks.
Incorrect Answers
A:Investigate disparate error rates, feature proxies and applicable employment nondiscrimination requirements. Not selected. It addresses a different governance mechanism, not the required assessment-and-control pair. It belongs with Employment discrimination: Equal treatment obligations apply to algorithmic employment decisions and proxy discrimination. The present case instead concerns Product safety liability.
D:Require human clearance and vendor-license review for commercial deliverables before release. Not selected. It addresses a different governance mechanism, not the required assessment-and-control pair. It belongs with Generated-output IP: AI output availability does not guarantee ownership, exclusivity or freedom from infringement. The present case instead concerns Product safety liability.
Question 393
A manufacturer embeds computer vision in a robotic safety stop but tests only a limited lighting condition. A supplier or jurisdictional change requires a fresh legal-control review. Which diagnostic task establishes the risk relevant to this described system?
- Identify foreseeable misuse, design defects, failure modes and relevant product safety duties.
- Investigate disparate error rates, feature proxies and applicable employment nondiscrimination requirements.
- Examine whether the feature yields disproportionate outcomes under applicable lending rules.
- Obtain rights or use lawfully eligible content and retain a traceable licensing inventory.
Explanation
Correct Answer
A:Identify foreseeable misuse, design defects, failure modes and relevant product safety duties. Correct. It establishes the relevant factual risk boundary before implementation. AI systems inside products can raise ordinary safety and product-liability risks.
Incorrect Answers
B:Investigate disparate error rates, feature proxies and applicable employment nondiscrimination requirements. Not selected. That check primarily diagnoses a different governance issue. It belongs with Employment discrimination: Equal treatment obligations apply to algorithmic employment decisions and proxy discrimination. The present case instead concerns Product safety liability.
C:Examine whether the feature yields disproportionate outcomes under applicable lending rules. Not selected. That check primarily diagnoses a different governance issue. It belongs with Credit and lending fairness: AI credit decisions remain subject to lending fairness, transparency and contestability obligations. The present case instead concerns Product safety liability.
D:Obtain rights or use lawfully eligible content and retain a traceable licensing inventory. Not selected. This action relates to Training-data copyright. Public accessibility does not automatically grant permission to reproduce or train on protected material. For the diagnose decision, it does not resolve the specifically described issue of Product safety liability in A manufacturer embeds computer vision in a robotic safety stop but tests only a limited lighting condition. A supplier or jurisdictional change requires a fresh legal-control review.
Question 394
A manufacturer embeds computer vision in a robotic safety stop but tests only a limited lighting condition. A supplier or jurisdictional change requires a fresh legal-control review. What is the strongest proportionate corrective design decision?
- Obtain rights or use lawfully eligible content and retain a traceable licensing inventory.
- Require human clearance and vendor-license review for commercial deliverables before release.
- Require validated fail-safe design, representative safety testing and defect-management controls.
- Remove unjustified proxy features and establish defensible credit-relevance and adverse-action review controls.
Explanation
Correct Answer
C:Require validated fail-safe design, representative safety testing and defect-management controls. Correct. It directly reduces or governs the specific documented risk. AI systems inside products can raise ordinary safety and product-liability risks.
Incorrect Answers
A:Obtain rights or use lawfully eligible content and retain a traceable licensing inventory. Not selected. That action treats another control objective, not the specific gap described. It belongs with Training-data copyright: Public accessibility does not automatically grant permission to reproduce or train on protected material. The present case instead concerns Product safety liability.
B:Require human clearance and vendor-license review for commercial deliverables before release. Not selected. That action treats another control objective, not the specific gap described. It belongs with Generated-output IP: AI output availability does not guarantee ownership, exclusivity or freedom from infringement. The present case instead concerns Product safety liability.
D:Remove unjustified proxy features and establish defensible credit-relevance and adverse-action review controls. Not selected. That action treats another control objective, not the specific gap described. It belongs with Credit and lending fairness: AI credit decisions remain subject to lending fairness, transparency and contestability obligations. The present case instead concerns Product safety liability.
Question 395
A manufacturer embeds computer vision in a robotic safety stop but tests only a limited lighting condition. A supplier or jurisdictional change requires a fresh legal-control review. Which observable evidence best demonstrates the chosen safeguard is working?
- Reevaluate relevant subgroup outcomes and lawful reason notices after the change.
- Sample training records to verify each restricted source has a documented permitted basis.
- Determine source licenses, access conditions, copyright exceptions and jurisdiction-specific permissions.
- Test hazardous edge cases and record whether safe-stop performance meets the accepted criterion.
Explanation
Correct Answer
D:Test hazardous edge cases and record whether safe-stop performance meets the accepted criterion. Correct. It yields measurable evidence that the case-specific safeguard works. AI systems inside products can raise ordinary safety and product-liability risks.
Incorrect Answers
A:Reevaluate relevant subgroup outcomes and lawful reason notices after the change. Not selected. That test would not demonstrate this particular risk has been addressed. It belongs with Credit and lending fairness: AI credit decisions remain subject to lending fairness, transparency and contestability obligations. The present case instead concerns Product safety liability.
B:Sample training records to verify each restricted source has a documented permitted basis. Not selected. That test would not demonstrate this particular risk has been addressed. It belongs with Training-data copyright: Public accessibility does not automatically grant permission to reproduce or train on protected material. The present case instead concerns Product safety liability.
C:Determine source licenses, access conditions, copyright exceptions and jurisdiction-specific permissions. Not selected. This action relates to Training-data copyright. Public accessibility does not automatically grant permission to reproduce or train on protected material. For the validate decision, it does not resolve the specifically described issue of Product safety liability in A manufacturer embeds computer vision in a robotic safety stop but tests only a limited lighting condition. A supplier or jurisdictional change requires a fresh legal-control review.
Question 396
A manufacturer embeds computer vision in a robotic safety stop but tests only a limited lighting condition. A supplier or jurisdictional change requires a fresh legal-control review. A material exception remains unresolved. Which containment decision best protects affected parties?
- Suspend automated adverse screening where evidence suggests discriminatory impact.
- Disable the unsafe autonomy mode and initiate product incident procedures if safety thresholds fail.
- Escalate discriminatory or unexplained denials for manual reassessment.
- Withdraw suspect assets and stop distribution while infringement risks are assessed.
Explanation
Correct Answer
B:Disable the unsafe autonomy mode and initiate product incident procedures if safety thresholds fail. Correct. It limits further reliance or exposure while the risk remains unresolved. AI systems inside products can raise ordinary safety and product-liability risks.
Incorrect Answers
A:Suspend automated adverse screening where evidence suggests discriminatory impact. Not selected. That response contains a different failure mode and leaves the stated risk unresolved. It belongs with Employment discrimination: Equal treatment obligations apply to algorithmic employment decisions and proxy discrimination. The present case instead concerns Product safety liability.
C:Escalate discriminatory or unexplained denials for manual reassessment. Not selected. That response contains a different failure mode and leaves the stated risk unresolved. It belongs with Credit and lending fairness: AI credit decisions remain subject to lending fairness, transparency and contestability obligations. The present case instead concerns Product safety liability.
D:Withdraw suspect assets and stop distribution while infringement risks are assessed. Not selected. That response contains a different failure mode and leaves the stated risk unresolved. It belongs with Generated-output IP: AI output availability does not guarantee ownership, exclusivity or freedom from infringement. The present case instead concerns Product safety liability.
Question 397
A provider labels a new employee-screening AI as minimal risk because its model is small and inexpensive. Legal and AI governance staff must agree on the defensible compliance response. Which diagnostic task establishes the risk relevant to this described system?
- Identify the role and applicable general-purpose model duties, including documentation and any systemic-risk obligations.
- Review gaps in the applicable quality-management system, responsibilities, monitoring and corrective actions.
- Classify the actual intended use, affected rights and prohibited or high-risk categories under the applicable EU framework.
- Map real modifications, branding and distribution activities to roles defined by applicable AI rules.
Explanation
Correct Answer
C:Classify the actual intended use, affected rights and prohibited or high-risk categories under the applicable EU framework. Correct. It establishes the relevant factual risk boundary before implementation. AI risk classification depends substantially on application and purpose, not computational scale.
Incorrect Answers
A:Identify the role and applicable general-purpose model duties, including documentation and any systemic-risk obligations. Not selected. That check primarily diagnoses a different governance issue. It belongs with General-purpose AI obligations: General-purpose models can carry model-level duties distinct from individual high-risk applications. The present case instead concerns EU AI Act risk classification.
B:Review gaps in the applicable quality-management system, responsibilities, monitoring and corrective actions. Not selected. That check primarily diagnoses a different governance issue. It belongs with High-risk quality management: High-risk AI compliance is a managed lifecycle obligation rather than a one-time test pass. The present case instead concerns EU AI Act risk classification.
D:Map real modifications, branding and distribution activities to roles defined by applicable AI rules. Not selected. That check primarily diagnoses a different governance issue. It belongs with AI legal role allocation: Legal AI roles may change with material modifications or new marketing claims. The present case instead concerns EU AI Act risk classification.
Question 398
A provider labels a new employee-screening AI as minimal risk because its model is small and inexpensive. Legal and AI governance staff must agree on the defensible compliance response. What is the strongest proportionate corrective design decision?
- Apply the resulting legal obligations to the use case instead of using model size as the risk category.
- Reassign and document legal duties where modifications or rebranding change the party's role.
- Provide meaningful human oversight, stop controls and trained reviewer authority appropriate to the law and risk.
- Provide clear contextual AI notices and escalation to a human when appropriate.
Explanation
Correct Answer
A:Apply the resulting legal obligations to the use case instead of using model size as the risk category. Correct. It directly reduces or governs the specific documented risk. AI risk classification depends substantially on application and purpose, not computational scale.
Incorrect Answers
B:Reassign and document legal duties where modifications or rebranding change the party's role. Not selected. That action treats another control objective, not the specific gap described. It belongs with AI legal role allocation: Legal AI roles may change with material modifications or new marketing claims. The present case instead concerns EU AI Act risk classification.
C:Provide meaningful human oversight, stop controls and trained reviewer authority appropriate to the law and risk. Not selected. That action treats another control objective, not the specific gap described. It belongs with Human oversight for high risk: A nominal human on the screen is not meaningful oversight without agency and usable controls. The present case instead concerns EU AI Act risk classification.
D:Provide clear contextual AI notices and escalation to a human when appropriate. Not selected. That action treats another control objective, not the specific gap described. It belongs with AI transparency notification: A transparent deployment should not mislead people about whether they are engaging with AI. The present case instead concerns EU AI Act risk classification.
Question 399
A provider labels a new employee-screening AI as minimal risk because its model is small and inexpensive. Legal and AI governance staff must agree on the defensible compliance response. Which observable evidence best demonstrates the chosen safeguard is working?
- Check a downstream integrator can identify capabilities, limits and intended safeguards from supplied documentation.
- Check the technical file links the chosen risk class to the actual product function and downstream deployment.
- Walk through an independent file review to confirm version, provenance, limits and risks are coherent.
- Test typical user journeys to confirm disclosures appear before misleading reliance.
Explanation
Correct Answer
B:Check the technical file links the chosen risk class to the actual product function and downstream deployment. Correct. It yields measurable evidence that the case-specific safeguard works. AI risk classification depends substantially on application and purpose, not computational scale.
Incorrect Answers
A:Check a downstream integrator can identify capabilities, limits and intended safeguards from supplied documentation. Not selected. That test would not demonstrate this particular risk has been addressed. It belongs with General-purpose AI obligations: General-purpose models can carry model-level duties distinct from individual high-risk applications. The present case instead concerns EU AI Act risk classification.
C:Walk through an independent file review to confirm version, provenance, limits and risks are coherent. Not selected. That test would not demonstrate this particular risk has been addressed. It belongs with EU technical documentation: Technical documentation supports conformity, regulator scrutiny and lifecycle accountability. The present case instead concerns EU AI Act risk classification.
D:Test typical user journeys to confirm disclosures appear before misleading reliance. Not selected. That test would not demonstrate this particular risk has been addressed. It belongs with AI transparency notification: A transparent deployment should not mislead people about whether they are engaging with AI. The present case instead concerns EU AI Act risk classification.
Question 400
A provider labels a new employee-screening AI as minimal risk because its model is small and inexpensive. Legal and AI governance staff must agree on the defensible compliance response. A material exception remains unresolved. Which containment decision best protects affected parties?
- Hold an unapproved high-risk model release until required quality evidence is available.
- Pause resale where the changed role has unmet compliance prerequisites.
- Block market release pending a correct classification and necessary safeguards.
- Restrict model release until applicable general-purpose obligations have been assessed and met.
Explanation
Correct Answer
C:Block market release pending a correct classification and necessary safeguards. Correct. It limits further reliance or exposure while the risk remains unresolved. AI risk classification depends substantially on application and purpose, not computational scale.
Incorrect Answers
A:Hold an unapproved high-risk model release until required quality evidence is available. Not selected. That response contains a different failure mode and leaves the stated risk unresolved. It belongs with High-risk quality management: High-risk AI compliance is a managed lifecycle obligation rather than a one-time test pass. The present case instead concerns EU AI Act risk classification.
B:Pause resale where the changed role has unmet compliance prerequisites. Not selected. That response contains a different failure mode and leaves the stated risk unresolved. It belongs with AI legal role allocation: Legal AI roles may change with material modifications or new marketing claims. The present case instead concerns EU AI Act risk classification.
D:Restrict model release until applicable general-purpose obligations have been assessed and met. Not selected. That response contains a different failure mode and leaves the stated risk unresolved. It belongs with General-purpose AI obligations: General-purpose models can carry model-level duties distinct from individual high-risk applications. The present case instead concerns EU AI Act risk classification.