IAPP AIGP Practice Test 63 - 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 63

Question 620

A model-hosting team monitors API uptime but never reviews model artifact tampering or dependency compromise. Which governance intervention most directly addresses this specific situation?

  1. Reassess risk, owners and residual exposure whenever material features or sources change.
  2. Investigate drift with data, product and operational experts before changing infrastructure.
  3. Threat-model the artifact, build pipeline and inference dependencies and add integrity controls.
  4. Create accurate technical and operational disclosures linked to the released model version.

Explanation

Correct Answer

C:

Threat-model the artifact, build pipeline and inference dependencies and add integrity controls. The action addresses threat modeling the model supply chain rather than treating every AI control as interchangeable. The decisive signal to verify is that the deployment accepts unsigned model files from an unverified build job. This matters because aI system security includes training and release assets, not just inference endpoints. Success should be measured by confirming that only approved, integrity-verified artifacts are promoted and alerts cover tampering attempts.

Incorrect Answers

A:

Reassess risk, owners and residual exposure whenever material features or sources change. AI system security includes training and release assets, not just inference endpoints. Threat-model the artifact, build pipeline and inference dependencies and add integrity controls.

B:

Investigate drift with data, product and operational experts before changing infrastructure. AI system security includes training and release assets, not just inference endpoints. Threat-model the artifact, build pipeline and inference dependencies and add integrity controls.

D:

Create accurate technical and operational disclosures linked to the released model version. AI system security includes training and release assets, not just inference endpoints. Threat-model the artifact, build pipeline and inference dependencies and add integrity controls.

Question 621

A retail forecast model abruptly degrades after a data pipeline changes units from dollars to cents. Which test result would best demonstrate that the actual governance concern has been addressed?

  1. Invalid batches are blocked and recovery replays the last accepted data version.
  2. Validated privacy tests support the release conditions and risky synthetic records are excluded.
  3. Post-validation probabilities are calibrated for the intended decision context or clearly limited.
  4. A third party can reproduce the released model's evaluation within declared tolerances.

Explanation

Correct Answer

A:

Invalid batches are blocked and recovery replays the last accepted data version. The required intervention is to add schema, range and distribution validation at ingestion with lineage and rollback. The particular diagnostic issue is that feature values jump by a factor of 100 without a corresponding real business change. The governing distinction is that data integrity checks are essential governance controls for model training and retraining.

Incorrect Answers

B:

Validated privacy tests support the release conditions and risky synthetic records are excluded. Data integrity checks are essential governance controls for model training and retraining. Invalid batches are blocked and recovery replays the last accepted data version.

C:

Post-validation probabilities are calibrated for the intended decision context or clearly limited. Data integrity checks are essential governance controls for model training and retraining. Invalid batches are blocked and recovery replays the last accepted data version.

D:

A third party can reproduce the released model's evaluation within declared tolerances. Data integrity checks are essential governance controls for model training and retraining. Invalid batches are blocked and recovery replays the last accepted data version.

Question 622

A demand predictor is retrained every Friday without checking whether data changed or validation improved. Which governance reasoning best supports the appropriate response?

  1. Runtime monitoring should connect measurable harms to intervention authority.
  2. Model documentation supports transparent, accountable transition from development to use.
  3. Retraining cadence should respond to model condition and risk rather than calendar habit alone.
  4. Issue management needs objective evidence of mitigation, not simply completion of a work item.

Explanation

Correct Answer

C:

Retraining cadence should respond to model condition and risk rather than calendar habit alone. This reasoning fits retraining decision schedule because the observed gap is that several weekly retrains degrade performance while the automation promotes each version. It points to the practical intervention: Use an evidence-driven maintenance schedule with material-change triggers and acceptance checks. The outcome should then show that only validated retrained models pass release gates; unneeded changes are skipped.

Incorrect Answers

A:

Runtime monitoring should connect measurable harms to intervention authority. This reasoning fits retraining decision schedule because the observed gap is that several weekly retrains degrade performance while the automation promotes each version. Retraining cadence should respond to model condition and risk rather than calendar habit alone.

B:

Model documentation supports transparent, accountable transition from development to use. This reasoning fits retraining decision schedule because the observed gap is that several weekly retrains degrade performance while the automation promotes each version. Retraining cadence should respond to model condition and risk rather than calendar habit alone.

D:

Issue management needs goal evidence of mitigation, not simply completion of a work item. This reasoning fits retraining decision schedule because the observed gap is that several weekly retrains degrade performance while the automation promotes each version. Retraining cadence should respond to model condition and risk rather than calendar habit alone.

Question 623

A hospital buys a medical triage model and assumes the vendor will also train clinicians and manage local patient complaints. Which governance intervention most directly addresses this specific situation?

  1. Train users on permitted workflows with role-specific examples, checks and escalation practice.
  2. Set central minimum controls with delegated regional review where laws and risks differ.
  3. Define vendor/provider duties separately from hospital deployment, monitoring, staff training and local escalation duties.
  4. Introduce a maintained AI inventory recording owners, purposes, providers, data and risk classifications.

Explanation

Correct Answer

C:

Define vendor/provider duties separately from hospital deployment, monitoring, staff training and local escalation duties. Building and using an AI system create different accountability obligations even when shared among organizations.

Incorrect Answers

A:

Train users on permitted workflows with role-specific examples, checks and escalation practice. Building and using an AI system create different accountability obligations even when shared among organizations. Define vendor/provider duties separately from hospital deployment, monitoring, staff training and local escalation duties.

B:

Set central minimum controls with delegated regional review where laws and risks differ. Building and using an AI system create different accountability obligations even when shared among organizations. Define vendor/provider duties separately from hospital deployment, monitoring, staff training and local escalation duties.

D:

Introduce a maintained AI inventory recording owners, purposes, providers, data and risk classifications. Building and using an AI system create different accountability obligations even when shared among organizations. Define vendor/provider duties separately from hospital deployment, monitoring, staff training and local escalation duties.

Question 624

A support chatbot inadvertently reveals a customer record in another customer's conversation, but operations classifies it solely as a model bug. Select the ONE intervention-and-verification pair that directly resolves this specific governance gap.

  1. Intervention: Clarify controller purpose decisions and processor contractual duties and permitted processing instructions. / Verification: Signed terms allocate data-protection roles and operational tests confirm instruction-bound processing.
  2. Intervention: Investigate a potential personal-data breach and apply the relevant notification and containment rules. / Verification: A documented lawful-purpose assessment determines permitted training use and any required additional notices or safeguards.
  3. Intervention: Restrict training features to relevant fields, justify any sensitive data use and reduce identifiability. / Verification: Incident procedures capture legal notification decisions, containment, evidence and correction.
  4. Intervention: Investigate a potential personal-data breach and apply the relevant notification and containment rules. / Verification: Incident procedures capture legal notification decisions, containment, evidence and correction.

Explanation

Correct Answer

D:

Intervention: Investigate a potential personal-data breach and apply the relevant notification and containment rules. / Verification: Incident procedures capture legal notification decisions, containment, evidence and correction.

Incorrect Answers

A:

Intervention: Clarify controller purpose decisions and processor contractual duties and permitted processing instructions. / Verification: Signed terms allocate data-protection roles and operational tests confirm instruction-bound processing. Intervention: Investigate a potential personal-data breach and apply the relevant notification and containment rules. / Verification: Incident procedures capture legal notification decisions, containment, evidence and correction.

B:

Intervention: Investigate a potential personal-data breach and apply the relevant notification and containment rules. / Verification: A documented lawful-purpose assessment determines permitted training use and any required additional notices or safeguards. Intervention: Investigate a potential personal-data breach and apply the relevant notification and containment rules. / Verification: Incident procedures capture legal notification decisions, containment, evidence and correction.

C:

Intervention: Restrict training features to relevant fields, justify any sensitive data use and reduce identifiability. / Verification: Incident procedures capture legal notification decisions, containment, evidence and correction. Intervention: Investigate a potential personal-data breach and apply the relevant notification and containment rules. / Verification: Incident procedures capture legal notification decisions, containment, evidence and correction.

Question 625

An architecture studio has no rule on whether employees may ask an AI tool to generate confidential client plans. Which additional observation would most strongly confirm the particular risk or control gap?

  1. The vendor can supply uptime commitments but no documentation of dataset sourcing or exclusions.
  2. The product backlog includes speed metrics but no user-harm or review criteria.
  3. Different teams send client-sensitive plans to public assistants without a documented authorization standard.
  4. A support ticket reveals requests are routed to an unreviewed fourth-party hosting service.

Explanation

Correct Answer

C:

Different teams send client-sensitive plans to public assistants without a documented authorization standard.

Incorrect Answers

A:

The vendor can supply uptime commitments but no documentation of dataset sourcing or exclusions. Different teams send client-sensitive plans to public assistants without a documented authorization standard.

B:

The product backlog includes speed metrics but no user-harm or review criteria. Different teams send client-sensitive plans to public assistants without a documented authorization standard.

D:

A support ticket reveals requests are routed to an unreviewed fourth-party hosting service. Different teams send client-sensitive plans to public assistants without a documented authorization standard.

Question 626

A health insurer's AI segments customers into price groups but management cannot demonstrate the absence of unlawful rating factors. Which test result would best demonstrate that the actual governance concern has been addressed?

  1. The insurer documents permitted rating factors and tests for prohibited proxy effects.
  2. Credit decisions satisfy applicable fairness and adverse-action explanation controls after testing.
  3. Approved terms and controls prevent unauthorized retention or competitive use of the design.
  4. The training-data register contains documented rights or excludes nonpermitted works.

Explanation

Correct Answer

A:

The insurer documents permitted rating factors and tests for prohibited proxy effects. The required intervention is to review insurance fairness rules and model factors with legal and actuarial oversight. The particular diagnostic issue is that pricing models rely on opaque correlations that cannot be linked to permitted risk variables. The governing distinction is that an insurance model's technical accuracy does not guarantee its rates are legally permissible.

Incorrect Answers

B:

Credit decisions satisfy applicable fairness and adverse-action explanation controls after testing. An insurance model's technical accuracy does not guarantee its rates are legally permissible. The insurer documents permitted rating factors and tests for prohibited proxy effects.

C:

Approved terms and controls prevent unauthorized retention or competitive use of the design. An insurance model's technical accuracy does not guarantee its rates are legally permissible. The insurer documents permitted rating factors and tests for prohibited proxy effects.

D:

The training-data register contains documented rights or excludes nonpermitted works. An insurance model's technical accuracy does not guarantee its rates are legally permissible. The insurer documents permitted rating factors and tests for prohibited proxy effects.

Question 627

A fintech lab runs a chatbot pilot on customer complaints without evaluating whether the tool may give financial advice. Select the ONE intervention-and-verification pair that directly resolves this specific governance gap.

  1. Intervention: Require pre-pilot use-case registration, risk classification and approval of safe testing boundaries. / Verification: Design verification tests user recourse and affected-group outcomes alongside technical objectives.
  2. Intervention: Define incident logging and evidence-retention procedures matched to data protection constraints. / Verification: A simulated incident produces bounded, access-controlled records sufficient to reconstruct cause.
  3. Intervention: Add provenance, lawful-use warranties, incident reporting and audit cooperation to the vendor review. / Verification: Every customer-facing experiment has a signed risk record and constrained pilot plan before access.
  4. Intervention: Require pre-pilot use-case registration, risk classification and approval of safe testing boundaries. / Verification: Every customer-facing experiment has a signed risk record and constrained pilot plan before access.

Explanation

Correct Answer

D:

Intervention: Require pre-pilot use-case registration, risk classification and approval of safe testing boundaries. / Verification: Every customer-facing experiment has a signed risk record and constrained pilot plan before access.

Incorrect Answers

A:

Intervention: Require pre-pilot use-case registration, risk classification and approval of safe testing boundaries. / Verification: Design verification tests user recourse and affected-group outcomes alongside technical goal. Intervention: Require pre-pilot use-case registration, risk classification and approval of safe testing boundaries. / Verification: Every customer-facing experiment has a signed risk record and constrained pilot plan before access.

B:

Intervention: Define incident logging and evidence-retention procedures matched to data protection constraints. / Verification: A simulated incident produces bounded, access-controlled records sufficient to reconstruct cause. Intervention: Require pre-pilot use-case registration, risk classification and approval of safe testing boundaries. / Verification: Every customer-facing experiment has a signed risk record and constrained pilot plan before access.

C:

Intervention: Add provenance, lawful-use warranties, incident reporting and audit cooperation to the vendor review. / Verification: Every customer-facing experiment has a signed risk record and constrained pilot plan before access. Intervention: Require pre-pilot use-case registration, risk classification and approval of safe testing boundaries. / Verification: Every customer-facing experiment has a signed risk record and constrained pilot plan before access.

Question 628

A public authority buys hosted inference with a contractual EU data-region promise but cannot verify where support logs go. Which additional observation would most strongly confirm the particular risk or control gap?

  1. The vendor changes scores in production without linking the change to a tested model version.
  2. The firm owns the data pipeline, selection logic and operating decisions but has no control owners.
  3. The terms grant use rights but explicitly exclude certain warranties and ownership assurances.
  4. Application inference stays in-region while debug traces with personal data are exported elsewhere.

Explanation

Correct Answer

D:

Application inference stays in-region while debug traces with personal data are exported elsewhere. The underlying governance reason is that data residency must cover operational logs and subcontractors, not only primary inference.

Incorrect Answers

A:

The vendor changes scores in production without linking the change to a tested model version. The underlying governance reason is that data residency must cover operational logs and subcontractors, not only primary inference. Application inference stays in-region while debug traces with personal data are exported elsewhere.

B:

The firm owns the data pipeline, selection logic and operating decisions but has no control owners. The underlying governance reason is that data residency must cover operational logs and subcontractors, not only primary inference. Application inference stays in-region while debug traces with personal data are exported elsewhere.

C:

The terms grant use rights but explicitly exclude certain warranties and ownership assurances. The underlying governance reason is that data residency must cover operational logs and subcontractors, not only primary inference. Application inference stays in-region while debug traces with personal data are exported elsewhere.

Question 629

A nonprofit wants a custom large-language model for volunteer scheduling despite limited annotated data, expertise and budget. Which governance reasoning best supports the appropriate response?

  1. AI model type should fit the data, output and consequence of the use case.
  2. Hosting architecture determines control over data access, retention and third-party exposure.
  3. Model scale should be justified by actual performance and operational constraints.
  4. Model ownership choices should reflect capability, risk appetite and lifecycle cost.

Explanation

Correct Answer

D:

Model ownership choices should reflect capability, risk appetite and lifecycle cost. This reasoning fits build versus buy decision because the observed gap is that the organization lacks model-training infrastructure and its task is a narrow scheduling-assistance workflow. It points to the practical intervention: Compare an existing governed service with in-house development on requirements, risk, support and total cost. The outcome should then show that a documented make-or-buy decision selects a sustainable service with required controls and ownership.

Incorrect Answers

A:

AI model type should fit the data, output and consequence of the use case. This reasoning fits build versus buy decision because the observed gap is that the organization lacks model-training infrastructure and its task is a narrow scheduling-assistance workflow. Model ownership choices should reflect capability, risk appetite and lifecycle cost.

B:

Hosting architecture determines control over data access, retention and third-party exposure. This reasoning fits build versus buy decision because the observed gap is that the organization lacks model-training infrastructure and its task is a narrow scheduling-assistance workflow. Model ownership choices should reflect capability, risk appetite and lifecycle cost.

C:

Model scale should be justified by actual performance and operational constraints. This reasoning fits build versus buy decision because the observed gap is that the organization lacks model-training infrastructure and its task is a narrow scheduling-assistance workflow. Model ownership choices should reflect capability, risk appetite and lifecycle cost.

Question 630

A hospital proposes a staff chatbot and discovers nurses paste full patient histories into an external tool. Which governance reasoning best supports the appropriate response?

  1. Fluent generative text can be unsupported; truthfulness demands verification rather than stylistic confidence.
  2. Agent autonomy changes the harm pathway because software can act, not merely recommend.
  3. AI interactions may themselves be a new personal-data collection and disclosure channel.
  4. Complex model accuracy does not remove the need for contextual explainability and accountability.

Explanation

Correct Answer

C:

AI interactions may themselves be a new personal-data collection and disclosure channel. This reasoning fits privacy in conversational prompts because the observed gap is that usage samples include directly identifying patient narratives unnecessary for the requested administrative summary. It points to the practical intervention: Limit permitted prompt data, restrict vendor retention and implement privacy-preserving input controls. The outcome should then show that pilot prompts exclude unnecessary identifiers and approved retention tests verify protected handling.

Incorrect Answers

A:

Fluent generative text can be unsupported; truthfulness demands verification rather than stylistic confidence. This reasoning fits privacy in conversational prompts because the observed gap is that usage samples include directly identifying patient narratives unnecessary for the requested administrative summary. AI interactions may themselves be a new personal-data collection and disclosure channel.

B:

Agent autonomy changes the harm pathway because software can act, not merely recommend. This reasoning fits privacy in conversational prompts because the observed gap is that usage samples include directly identifying patient narratives unnecessary for the requested administrative summary. AI interactions may themselves be a new personal-data collection and disclosure channel.

D:

Complex model accuracy does not remove the need for contextual explainability and accountability. This reasoning fits privacy in conversational prompts because the observed gap is that usage samples include directly identifying patient narratives unnecessary for the requested administrative summary. AI interactions may themselves be a new personal-data collection and disclosure channel.

Artificial Intelligence Governance Professional (AIGP) Practice Tests