IAPP AIGP Practice Test 18 - Prepaway
Practice Exams:

Pass IAPP AIGP Exam in First Attempt Guaranteed!

Get 100% Latest Exam Questions, Accurate & Verified Answers!

30 Days Free Updates, Instant Download!

Verified By Experts
Premium Bundle
$29.98
$79.99 $109.97

AIGP Premium Bundle


  • Premium File 222 Questions & Answers. Last update: Nov 15, 2026
  • Training Course 124 Lectures
  • Study Guide 566 Pages

IAPP AIGP Practice Test 18

Question 170

A manufacturing employer pilots worker-productivity AI without asking employees how shift patterns affect measurements. Which test result would best demonstrate that the actual governance concern has been addressed?

  1. Scenario-based training checks demonstrate staff choose approved tools and redact restricted information.
  2. The review documents shift exceptions and worker feedback with tested mitigation before use.
  3. A role matrix and contract assign each model-level and use-site control to a responsible party.
  4. Governance reports show a clear risk owner and decisions logged at the appropriate approval level.

Explanation

Correct Answer

B:

The review documents shift exceptions and worker feedback with tested mitigation before use. Affected stakeholder perspectives are substantive design evidence, not merely a communications exercise.

Incorrect Answers

A:

Scenario-based training checks demonstrate staff choose approved tools and redact restricted information. Affected stakeholder perspectives are substantive design evidence, not merely a communications exercise. The review documents shift exceptions and worker feedback with tested mitigation before use.

C:

A role matrix and contract assign each model-level and use-site control to a responsible party. Affected stakeholder perspectives are substantive design evidence, not merely a communications exercise. The review documents shift exceptions and worker feedback with tested mitigation before use.

D:

Governance reports show a clear risk owner and decisions logged at the appropriate approval level. Affected stakeholder perspectives are substantive design evidence, not merely a communications exercise. The review documents shift exceptions and worker feedback with tested mitigation before use.

Question 171

A job-matching AI passed its launch fairness test but applicants in a new region receive poorer recommendations. Which governance reasoning best supports the appropriate response?

  1. AI system security includes training and release assets, not just inference endpoints.
  2. AI model performance depends on a changing operational context, not just code stability.
  3. Responsible AI validation is continuous because populations and workflows change.
  4. Retraining cadence should respond to model condition and risk rather than calendar habit alone.

Explanation

Correct Answer

C:

Responsible AI validation is continuous because populations and workflows change. This reasoning fits fairness monitoring after release because the observed gap is that new-region outcomes deviate from tested group performance without a formal signal. It points to the practical intervention: Monitor subgroup outcomes after launch and route significant discrepancies to accountable review. The outcome should then show that a monitoring dashboard and remediation protocol detect and address material group differences.

Incorrect Answers

A:

AI system security includes training and release assets, not just inference endpoints. This reasoning fits fairness monitoring after release because the observed gap is that new-region outcomes deviate from tested group performance without a formal signal. Responsible AI validation is continuous because populations and workflows change.

B:

AI model performance depends on a changing operational context, not just code stability. This reasoning fits fairness monitoring after release because the observed gap is that new-region outcomes deviate from tested group performance without a formal signal. Responsible AI validation is continuous because populations and workflows change.

D:

Retraining cadence should respond to model condition and risk rather than calendar habit alone. This reasoning fits fairness monitoring after release because the observed gap is that new-region outcomes deviate from tested group performance without a formal signal. Responsible AI validation is continuous because populations and workflows change.

Question 172

A financial fraud model drifts, but the infrastructure team keeps restarting servers despite a silent label-definition change. Which governance reasoning best supports the appropriate response?

  1. Responsible AI validation is continuous because populations and workflows change.
  2. AI system security includes training and release assets, not just inference endpoints.
  3. Transparency requires actionable deployment information, not just generic AI branding.
  4. AI failures can arise from data and human processes even when compute is operational.

Explanation

Correct Answer

D:

AI failures can arise from data and human processes even when compute is operational. This reasoning fits cross-functional root cause because the observed gap is that servers stay healthy while label semantics changed in the reporting pipeline. It points to the practical intervention: Investigate drift with data, product and operational experts before changing infrastructure. The outcome should then show that a corrected label mapping and evaluation restore model reliability without unnecessary restarts.

Incorrect Answers

A:

Responsible AI validation is continuous because populations and workflows change. The outcome should then show that a corrected label mapping and evaluation restore model reliability without unnecessary restarts. AI failures can arise from data and human processes even when compute is operational.

B:

AI system security includes training and release assets, not just inference endpoints. The outcome should then show that a corrected label mapping and evaluation restore model reliability without unnecessary restarts. AI failures can arise from data and human processes even when compute is operational.

C:

Transparency requires actionable deployment information, not just generic AI branding. The outcome should then show that a corrected label mapping and evaluation restore model reliability without unnecessary restarts. AI failures can arise from data and human processes even when compute is operational.

Question 173

A policy team asks whether its trustworthy-AI approach should address human rights, transparency and accountability beyond technical safety. Which governance reasoning best supports the appropriate response?

  1. An AI impact assessment considers consequences beyond a model's accuracy score.
  2. The OECD principles offer high-level trustworthy-AI guidance rather than a substitute for binding law.
  3. NIST Manage translates risk understanding and measurement into action.
  4. NIST's Govern is a cross-cutting function informing Map, Measure and Manage.

Explanation

Correct Answer

B:

The OECD principles offer high-level trustworthy-AI guidance rather than a substitute for binding law. This reasoning fits oecd trustworthy ai principles because the observed gap is that the drafted ethics policy lists model accuracy but omits transparency and human-centered values. It points to the practical intervention: Map governance goals to the OECD AI Principles and translate broad principles into local controls. The outcome should then show that adopted policy connects trustworthy-AI principles to reviewable organizational practices.

Incorrect Answers

A:

An AI impact assessment considers consequences beyond a model's accuracy score. This reasoning fits oecd trustworthy ai principles because the observed gap is that the drafted ethics policy lists model accuracy but omits transparency and human-centered values. The OECD principles offer high-level trustworthy-AI guidance rather than a substitute for binding law.

C:

NIST Manage translates risk understanding and measurement into action. This reasoning fits oecd trustworthy ai principles because the observed gap is that the drafted ethics policy lists model accuracy but omits transparency and human-centered values. The OECD principles offer high-level trustworthy-AI guidance rather than a substitute for binding law.

D:

NIST's Govern is a cross-cutting function informing Map, Measure and Manage. This reasoning fits oecd trustworthy ai principles because the observed gap is that the drafted ethics policy lists model accuracy but omits transparency and human-centered values. The OECD principles offer high-level trustworthy-AI guidance rather than a substitute for binding law.

Question 174

An HR team considers sending confidential grievance documents to a public API without clarifying provider retention. Which governance reasoning best supports the appropriate response?

  1. Input modalities define what a model can observe and therefore whether it fits the task.
  2. Deployment placement should reflect latency, connectivity, safety and maintenance risks.
  3. Open availability and commercial purchase each create distinct governance obligations.
  4. Hosting architecture determines control over data access, retention and third-party exposure.

Explanation

Correct Answer

D:

Hosting architecture determines control over data access, retention and third-party exposure. This reasoning fits cloud hosted inference confidentiality because the observed gap is that the proposed cloud route retains inputs beyond the organization's authorized processing terms. It points to the practical intervention: Compare hosted processing with approved isolated alternatives and require contractual and technical safeguards. The outcome should then show that data-flow and vendor reviews verify confidential text stays within approved boundaries.

Incorrect Answers

A:

Input modalities define what a model can observe and therefore whether it fits the task. This reasoning fits cloud hosted inference confidentiality because the observed gap is that the proposed cloud route retains inputs beyond the organization's authorized processing terms. Hosting architecture determines control over data access, retention and third-party exposure.

B:

Deployment placement should reflect latency, connectivity, safety and maintenance risks. This reasoning fits cloud hosted inference confidentiality because the observed gap is that the proposed cloud route retains inputs beyond the organization's authorized processing terms. Hosting architecture determines control over data access, retention and third-party exposure.

C:

Open availability and commercial purchase each create distinct governance obligations. This reasoning fits cloud hosted inference confidentiality because the observed gap is that the proposed cloud route retains inputs beyond the organization's authorized processing terms. Hosting architecture determines control over data access, retention and third-party exposure.

Question 175

A customer-service team wants an assistant to follow a stable domain-specific classification rubric across thousands of cases. Which test result would best demonstrate that the actual governance concern has been addressed?

  1. Data-flow and vendor reviews verify confidential text stays within approved boundaries.
  2. A chosen adaptation approach improves rubric compliance on unseen test cases without harming safety.
  3. Validated predictions support maintenance planning with measurable false-alarm and miss rates.
  4. Offline field trials satisfy task metrics without unnecessary remote processing.

Explanation

Correct Answer

B:

A chosen adaptation approach improves rubric compliance on unseen test cases without harming safety. The required intervention is to compare prompt-based controls and fine-tuning on data rights, evaluations and operational cost. The particular diagnostic issue is that prompt-only tests remain inconsistent across representative labeled examples. The governing distinction is that model adaptation choices require evidence of task fit, data suitability and regression risk.

Incorrect Answers

A:

Data-flow and vendor reviews verify confidential text stays within approved boundaries. Model adaptation choices require evidence of task fit, data suitability and regression risk. A chosen adaptation approach improves rubric compliance on unseen test cases without harming safety.

C:

Validated predictions support maintenance planning with measurable false-alarm and miss rates. Model adaptation choices require evidence of task fit, data suitability and regression risk. A chosen adaptation approach improves rubric compliance on unseen test cases without harming safety.

D:

Offline field trials satisfy task metrics without unnecessary remote processing. Model adaptation choices require evidence of task fit, data suitability and regression risk. A chosen adaptation approach improves rubric compliance on unseen test cases without harming safety.

Question 176

A bank specifies only overall fraud-detection accuracy for a model affecting customers from several demographic groups. Which governance reasoning best supports the appropriate response?

  1. Traceability makes design choices accountable and supports change impact analysis.
  2. Responsible design requires system-level controls around a model, not reliance on prompt instructions alone.
  3. Fairness is a measurable design requirement where unequal errors affect people.
  4. Stronger preventive controls are generally preferable to user notices as sole risk mitigation.

Explanation

Correct Answer

C:

Fairness is a measurable design requirement where unequal errors affect people. This reasoning fits requirements for protected-group performance because the observed gap is that the requirements document hides false-positive differences by reporting one aggregate statistic. It points to the practical intervention: Add subgroup error thresholds and a remediation plan to the design acceptance criteria. The outcome should then show that signed acceptance tests cover relevant group performance and meaningful review of flagged accounts.

Incorrect Answers

A:

Traceability makes design choices accountable and supports change impact analysis. This reasoning fits requirements for protected-group performance because the observed gap is that the requirements document hides false-positive differences by reporting one aggregate statistic. Fairness is a measurable design requirement where unequal errors affect people.

B:

Responsible design requires system-level controls around a model, not reliance on prompt instructions alone. This reasoning fits requirements for protected-group performance because the observed gap is that the requirements document hides false-positive differences by reporting one aggregate statistic. Fairness is a measurable design requirement where unequal errors affect people.

D:

Stronger preventive controls are generally preferable to user notices as sole risk mitigation. This reasoning fits requirements for protected-group performance because the observed gap is that the requirements document hides false-positive differences by reporting one aggregate statistic. Fairness is a measurable design requirement where unequal errors affect people.

Question 177

A legal-aid nonprofit pilots text generation that cites nonexistent local regulations as grounds to deny assistance. Which governance reasoning best supports the appropriate response?

  1. Training and evaluation data distribution can create unequal AI performance at scale.
  2. Aggregate accuracy can hide discriminatory patterns and must be assessed against affected groups.
  3. Fluent generative text can be unsupported; truthfulness demands verification rather than stylistic confidence.
  4. Probabilistic predictions require empirical validation and fallback controls, unlike a deterministic policy lookup.

Explanation

Correct Answer

C:

Fluent generative text can be unsupported; truthfulness demands verification rather than stylistic confidence. This reasoning fits generative hallucination in advice because the observed gap is that a sampled response names an invented regulation that cannot be found in the authoritative legal database.

Incorrect Answers

A:

Training and evaluation data distribution can create unequal AI performance at scale. This reasoning fits generative hallucination in advice because the observed gap is that a sampled response names an invented regulation that cannot be found in the authoritative legal database. Fluent generative text can be unsupported; truthfulness demands verification rather than stylistic confidence.

B:

Aggregate accuracy can hide discriminatory patterns and must be assessed against affected groups. This reasoning fits generative hallucination in advice because the observed gap is that a sampled response names an invented regulation that cannot be found in the authoritative legal database. Fluent generative text can be unsupported; truthfulness demands verification rather than stylistic confidence.

D:

Probabilistic predictions require empirical validation and fallback controls, unlike a deterministic policy lookup. This reasoning fits generative hallucination in advice because the observed gap is that a sampled response names an invented regulation that cannot be found in the authoritative legal database. Fluent generative text can be unsupported; truthfulness demands verification rather than stylistic confidence.

Question 178

A benefits provider drafts a welfare prediction feature based solely on throughput targets, leaving out appeal rights. Which governance intervention most directly addresses this specific situation?

  1. Write response thresholds, emergency suspension roles and fallback service procedures.
  2. Create an AI acceptable-use policy describing allowed data, tool approvals and professional review.
  3. Add approved-subprocessor disclosure, flow-down controls and change notification to procurement processes.
  4. Add fairness, accessibility and contestability acceptance criteria to design requirements.

Explanation

Correct Answer

D:

Add fairness, accessibility and contestability acceptance criteria to design requirements. Responsible AI requirements belong in the product design, not only in a final release memo.

Incorrect Answers

A:

Write response thresholds, emergency suspension roles and fallback service procedures. Responsible AI requirements belong in the product design, not only in a final release memo. Add fairness, accessibility and contestability acceptance criteria to design requirements.

B:

Create an AI acceptable-use policy describing allowed data, tool approvals and professional review. Responsible AI requirements belong in the product design, not only in a final release memo. Add fairness, accessibility and contestability acceptance criteria to design requirements.

C:

Add approved-subprocessor disclosure, flow-down controls and change notification to procurement processes. Responsible AI requirements belong in the product design, not only in a final release memo. Add fairness, accessibility and contestability acceptance criteria to design requirements.

Question 179

A benefits agency deploys fraud screening without a simple channel for beneficiaries to dispute incorrect flags. Which governance intervention most directly addresses this specific situation?

  1. Implement jurisdiction-aware controls to suspend the affected use while preserving permitted deployments.
  2. Establish an accessible appeal and human case review route with recorded outcomes.
  3. Define performance drift triggers for targeted evaluation and approved retraining or suspension.
  4. Review and revoke AI tool identities, permissions and data scopes on a scheduled and event-driven basis.

Explanation

Correct Answer

B:

Establish an accessible appeal and human case review route with recorded outcomes. The action addresses complaint and appeal route rather than treating every AI control as interchangeable. The decisive signal to verify is that affected residents cannot correct their information or learn why assistance is delayed. This matters because contestability helps protect individuals when AI informs consequential decisions. Success should be measured by confirming that a test appeal reaches an authorized reviewer and resolves documented factual errors.

Incorrect Answers

A:

Implement jurisdiction-aware controls to suspend the affected use while preserving permitted deployments. Contestability helps protect individuals when AI informs consequential decisions. Establish an accessible appeal and human case review route with recorded outcomes.

C:

Define performance drift triggers for targeted evaluation and approved retraining or suspension. Contestability helps protect individuals when AI informs consequential decisions. Establish an accessible appeal and human case review route with recorded outcomes.

D:

Review and revoke AI tool identities, permissions and data scopes on a scheduled and event-driven basis. Contestability helps protect individuals when AI informs consequential decisions. Establish an accessible appeal and human case review route with recorded outcomes.

Question 180

A medical-device distributor collects patient complaints about its AI feature but does not link them to model versions or corrective measures. Which test result would best demonstrate that the actual governance concern has been addressed?

  1. A controlled test shows substantive reviewer decisions can change the outcome before final action.
  2. Observed adoption shifts toward an approved confidential process and sensitive uploads decline.
  3. Governance reviews can trace trends to models and verify corrective actions and disclosures.
  4. Tests show the restricted region is blocked without disabling unrelated lawful uses.

Explanation

Correct Answer

C:

Governance reviews can trace trends to models and verify corrective actions and disclosures. The required intervention is to maintain post-market evidence linking complaints, versions, trends and follow-up decisions. The particular diagnostic issue is that complaint tickets contain patient concerns with no product version or assessed severity. The governing distinction is that deployment oversight continues after sale and must preserve usable field evidence.

Incorrect Answers

A:

A controlled test shows substantive reviewer decisions can change the outcome before final action. Deployment oversight continues after sale and must preserve usable field evidence. Governance reviews can trace trends to models and verify corrective actions and disclosures.

B:

Observed adoption shifts toward an approved confidential process and sensitive uploads decline. Deployment oversight continues after sale and must preserve usable field evidence. Governance reviews can trace trends to models and verify corrective actions and disclosures.

D:

Tests show the restricted region is blocked without disabling unrelated lawful uses. Deployment oversight continues after sale and must preserve usable field evidence. Governance reviews can trace trends to models and verify corrective actions and disclosures.

Artificial Intelligence Governance Professional (AIGP) Practice Tests