IAPP AIGP Practice Test 6 - 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 6

Question 50

A service organization deploys a public chatbot whose interface impersonates a human employee without notice. Which additional observation would most strongly confirm the particular risk or control gap?

  1. A design memo targets identified vulnerabilities rather than legitimate customer assistance.
  2. Usability testing shows users reasonably believe they are speaking with a named human agent.
  3. The team cannot produce an inventory, model history or reportable incident record on request.
  4. An auditor cannot identify the tested version or its risk-control evidence.

Explanation

Correct Answer

B:

Usability testing shows users reasonably believe they are speaking with a named human agent.

Incorrect Answers

A:

A design memo targets identified vulnerabilities rather than legitimate customer assistance. Usability testing shows users reasonably believe they are speaking with a named human agent.

C:

The team cannot produce an inventory, model history or reportable incident record on request. Usability testing shows users reasonably believe they are speaking with a named human agent.

D:

An auditor cannot identify the tested version or its risk-control evidence. Usability testing shows users reasonably believe they are speaking with a named human agent.

Question 51

A fraud model transforms dozens of transaction fields but development loses the mapping between engineered features and raw records. Select the ONE intervention-and-verification pair that directly resolves this specific governance gap.

  1. Intervention: Build auditable data lineage from sources through transformations, splits and model versions. / Verification: Reviewers can trace the disputed score input to versioned transformation and source provenance.
  2. Intervention: Test untrusted retrieved instructions and prevent them from overriding system policy or tool permissions. / Verification: Reviewers can trace the disputed score input to versioned transformation and source provenance.
  3. Intervention: Build auditable data lineage from sources through transformations, splits and model versions. / Verification: A leakage-free holdout evaluation reports realistic generalization to unseen applicants.
  4. Intervention: Map the request to affected datasets and determine required deletion or retraining actions under applicable obligations. / Verification: The organization can document how the request was handled across relevant data and model artifacts.

Explanation

Correct Answer

A:

Intervention: Build auditable data lineage from sources through transformations, splits and model versions. / Verification: Reviewers can trace the disputed score input to versioned transformation and source provenance. The reason to require this combination is that data provenance supports reproducibility, investigation and accountability across AI pipelines.

Incorrect Answers

B:

Intervention: Test untrusted retrieved instructions and prevent them from overriding system policy or tool permissions. / Verification: Reviewers can trace the disputed score input to versioned transformation and source provenance. The reason to require this combination is that data provenance supports reproducibility, investigation and accountability across AI pipelines. Intervention: Build auditable data lineage from sources through transformations, splits and model versions. / Verification: Reviewers can trace the disputed score input to versioned transformation and source provenance.

C:

Intervention: Build auditable data lineage from sources through transformations, splits and model versions. / Verification: A leakage-free holdout evaluation reports realistic generalization to unseen applicants. The reason to require this combination is that data provenance supports reproducibility, investigation and accountability across AI pipelines. Intervention: Build auditable data lineage from sources through transformations, splits and model versions. / Verification: Reviewers can trace the disputed score input to versioned transformation and source provenance.

D:

Intervention: Map the request to affected datasets and determine required deletion or retraining actions under applicable obligations. / Verification: The organization can document how the request was handled across relevant data and model artifacts. The reason to require this combination is that data provenance supports reproducibility, investigation and accountability across AI pipelines. Intervention: Build auditable data lineage from sources through transformations, splits and model versions. / Verification: Reviewers can trace the disputed score input to versioned transformation and source provenance.

Question 52

A public health authority hires a cloud vendor to process survey text but leaves instructions and audit rights undefined. Select the ONE intervention-and-verification pair that directly resolves this specific governance gap.

  1. Intervention: Implement a process to propagate validated corrections and handle relevant data-subject requests in downstream AI assets. / Verification: Signed terms allocate data-protection roles and operational tests confirm instruction-bound processing.
  2. Intervention: Assess sensitive biometric processing restrictions and less intrusive access alternatives. / Verification: An approved assessment justifies processing conditions and demonstrates proportionate secure controls.
  3. 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.
  4. Intervention: Clarify controller purpose decisions and processor contractual duties and permitted processing instructions. / Verification: Authorized users retrieve only their entitled matter content in negative and positive access tests.

Explanation

Correct Answer

C:

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. The reason to require this combination is that outsourcing computation does not remove the controller's data-protection accountability.

Incorrect Answers

A:

Intervention: Implement a process to propagate validated corrections and handle relevant data-subject requests in downstream AI assets. / Verification: Signed terms allocate data-protection roles and operational tests confirm instruction-bound processing. The reason to require this combination is that outsourcing computation does not remove the controller's data-protection accountability. 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.

B:

Intervention: Assess sensitive biometric processing restrictions and less intrusive access alternatives. / Verification: An approved assessment justifies processing conditions and demonstrates proportionate secure controls. The reason to require this combination is that outsourcing computation does not remove the controller's data-protection accountability. 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.

D:

Intervention: Clarify controller purpose decisions and processor contractual duties and permitted processing instructions. / Verification: Authorized users retrieve only their entitled matter content in negative and positive access tests. The reason to require this combination is that outsourcing computation does not remove the controller's data-protection accountability. 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.

Question 53

An open-feedback classifier accepts untrusted public labels directly into an automated retraining pipeline. Which governance intervention most directly addresses this specific situation?

  1. Compare candidate architectures against task complexity, interpretability, cost and evidence requirements.
  2. Rate harms by impact and likelihood with explicit tolerance and escalation thresholds.
  3. Threat-model malicious training input and design validation and isolation controls before ingestion.
  4. Challenge the need for AI and compare the simpler rules-based alternative before approving design.

Explanation

Correct Answer

C:

Threat-model malicious training input and design validation and isolation controls before ingestion. The action addresses data-poisoning threat model at design rather than treating every AI control as interchangeable. The decisive signal to verify is that the proposed pipeline treats anonymous labels as authoritative without provenance checks. This matters because aI threat models must include attacks on data and model behavior, not only network intrusion. Success should be measured by confirming that adversarial inputs are quarantined or rejected before they affect approved training assets.

Incorrect Answers

A:

Compare candidate architectures against task complexity, interpretability, cost and evidence requirements. AI threat models must include attacks on data and model behavior, not only network intrusion. Threat-model malicious training input and design validation and isolation controls before ingestion.

B:

Rate harms by impact and likelihood with explicit tolerance and escalation thresholds. AI threat models must include attacks on data and model behavior, not only network intrusion. Threat-model malicious training input and design validation and isolation controls before ingestion.

D:

Challenge the need for AI and compare the simpler rules-based alternative before approving design. AI threat models must include attacks on data and model behavior, not only network intrusion. Threat-model malicious training input and design validation and isolation controls before ingestion.

Question 54

A global shipping firm distributes a one-page AI policy but frontline dispatchers keep uploading live manifests to unsanctioned chatbots. Which governance reasoning best supports the appropriate response?

  1. Roles in an AI supply chain are not interchangeable; each controls a different part of the risk.
  2. A policy without practical role-based understanding will not consistently change behavior.
  3. Federated AI governance can preserve core standards without erasing local responsibilities.
  4. Effective AI oversight needs assigned responsibility rather than a committee name without authority.

Explanation

Correct Answer

B:

A policy without practical role-based understanding will not consistently change behavior. This reasoning fits targeted staff awareness because the observed gap is that incident reports show staff misunderstand which shipment fields may enter third-party AI services. It points to the practical intervention: Train users on permitted workflows with role-specific examples, checks and escalation practice. The outcome should then show that scenario-based training checks demonstrate staff choose approved tools and redact restricted information.

Incorrect Answers

A:

Roles in an AI supply chain are not interchangeable; each controls a different part of the risk. This reasoning fits targeted staff awareness because the observed gap is that incident reports show staff misunderstand which shipment fields may enter third-party AI services. A policy without practical role-based understanding will not consistently change behavior.

C:

Federated AI governance can preserve core standards without erasing local responsibilities. This reasoning fits targeted staff awareness because the observed gap is that incident reports show staff misunderstand which shipment fields may enter third-party AI services. A policy without practical role-based understanding will not consistently change behavior.

D:

Effective AI oversight needs assigned responsibility rather than a committee name without authority. This reasoning fits targeted staff awareness because the observed gap is that incident reports show staff misunderstand which shipment fields may enter third-party AI services. A policy without practical role-based understanding will not consistently change behavior.

Question 55

A company builds a proprietary resume screening model and believes avoiding an external vendor removes responsibility for harm. Which test result would best demonstrate that the actual governance concern has been addressed?

  1. A simulated incident triggers agreed notification and preservation of relevant evidence.
  2. Legal approval records the actual allocation of claim risk for the publishing workflow.
  3. The risk register assigns internal design and deployment obligations with evidence of compliance.
  4. The contract and assurance process provide defensible evidence for the bank's oversight duties.

Explanation

Correct Answer

C:

The risk register assigns internal design and deployment obligations with evidence of compliance. The required intervention is to assess direct developer and deployer responsibilities, documentation and liability exposure. The particular diagnostic issue is that the firm owns the data pipeline, selection logic and operating decisions but has no control owners. The governing distinction is that building AI in-house can increase direct accountability rather than eliminate risk.

Incorrect Answers

A:

A simulated incident triggers agreed notification and preservation of relevant evidence. Building AI in-house can increase direct accountability rather than eliminate risk. The risk register assigns internal design and deployment obligations with evidence of compliance.

B:

Legal approval records the actual allocation of claim risk for the publishing workflow. Building AI in-house can increase direct accountability rather than eliminate risk. The risk register assigns internal design and deployment obligations with evidence of compliance.

D:

The contract and assurance process provide defensible evidence for the bank's oversight duties. Building AI in-house can increase direct accountability rather than eliminate risk. The risk register assigns internal design and deployment obligations with evidence of compliance.

Question 56

A company acquires an analytics vendor whose data use policy conflicts with the buyer's AI standards. Which additional observation would most strongly confirm the particular risk or control gap?

  1. Purchase approval contains cost forecasts but no evidence of model behavior or fairness review.
  2. Operations staff can see an alert but no one has authority to halt automated recommendations.
  3. The product backlog includes speed metrics but no user-harm or review criteria.
  4. Two overlapping retention and consent requirements apply with no governing decision.

Explanation

Correct Answer

D:

Two overlapping retention and consent requirements apply with no governing decision.

Incorrect Answers

A:

Purchase approval contains cost forecasts but no evidence of model behavior or fairness review. Two overlapping retention and consent requirements apply with no governing decision.

B:

Operations staff can see an alert but no one has authority to halt automated recommendations. Two overlapping retention and consent requirements apply with no governing decision.

C:

The product backlog includes speed metrics but no user-harm or review criteria. Two overlapping retention and consent requirements apply with no governing decision.

Question 57

A public-service translation model is trained mainly on urban speech and performs poorly for rural dialect speakers. Which test result would best demonstrate that the actual governance concern has been addressed?

  1. Evaluation demonstrates acceptable subgroup coverage and performance against defined criteria.
  2. Validated privacy tests support the release conditions and risky synthetic records are excluded.
  3. Invalid batches are blocked and recovery replays the last accepted data version.
  4. A third party can reproduce the released model's evaluation within declared tolerances.

Explanation

Correct Answer

A:

Evaluation demonstrates acceptable subgroup coverage and performance against defined criteria. Training distribution should represent the deployment population rather than the easiest data source.

Incorrect Answers

B:

Validated privacy tests support the release conditions and risky synthetic records are excluded. Training distribution should represent the deployment population rather than the easiest data source. Evaluation demonstrates acceptable subgroup coverage and performance against defined criteria.

C:

Invalid batches are blocked and recovery replays the last accepted data version. Training distribution should represent the deployment population rather than the easiest data source. Evaluation demonstrates acceptable subgroup coverage and performance against defined criteria.

D:

A third party can reproduce the released model's evaluation within declared tolerances. Training distribution should represent the deployment population rather than the easiest data source. Evaluation demonstrates acceptable subgroup coverage and performance against defined criteria.

Question 58

A public-service translation model is trained mainly on urban speech and performs poorly for rural dialect speakers. Which additional observation would most strongly confirm the particular risk or control gap?

  1. Dataset profiling reveals that rural utterances represent a tiny share of the intended population.
  2. Two analysts obtain incompatible results despite citing the same report title.
  3. Nearest-neighbor checks show several generated records nearly match uniquely identifying real loans.
  4. Feature values jump by a factor of 100 without a corresponding real business change.

Explanation

Correct Answer

A:

Dataset profiling reveals that rural utterances represent a tiny share of the intended population. The underlying governance reason is that training distribution should represent the deployment population rather than the easiest data source.

Incorrect Answers

B:

Two analysts obtain incompatible results despite citing the same report title. The underlying governance reason is that training distribution should represent the deployment population rather than the easiest data source. Dataset profiling reveals that rural utterances represent a tiny share of the intended population.

C:

Nearest-neighbor checks show several generated records nearly match uniquely identifying real loans. The underlying governance reason is that training distribution should represent the deployment population rather than the easiest data source. Dataset profiling reveals that rural utterances represent a tiny share of the intended population.

D:

Feature values jump by a factor of 100 without a corresponding real business change. The underlying governance reason is that training distribution should represent the deployment population rather than the easiest data source. Dataset profiling reveals that rural utterances represent a tiny share of the intended population.

Question 59

An education company procures a language model with unclear training-data rights and only negotiates availability service levels. 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. A new version materially alters rankings while the release record only shows a deployment timestamp.
  3. The trial has live user access but no documented intended use, guardrails or risk owner.
  4. Two overlapping retention and consent requirements apply with no governing decision.

Explanation

Correct Answer

A:

The vendor can supply uptime commitments but no documentation of dataset sourcing or exclusions.

Incorrect Answers

B:

A new version materially alters rankings while the release record only shows a deployment timestamp. The vendor can supply uptime commitments but no documentation of dataset sourcing or exclusions.

C:

The trial has live user access but no documented intended use, guardrails or risk owner. The vendor can supply uptime commitments but no documentation of dataset sourcing or exclusions.

D:

Two overlapping retention and consent requirements apply with no governing decision. The vendor can supply uptime commitments but no documentation of dataset sourcing or exclusions.

Question 60

A public agency names an employee as a reviewer but does not give them authority to challenge the AI decision. Which governance intervention most directly addresses this specific situation?

  1. Assess whether the employment selection use falls within a high-risk AI category and plan controls under applicable timelines.
  2. Evaluate the applicable GPAI provider duties for technical and downstream documentation.
  3. Design human oversight with comprehension, time, access to evidence and power to override where required.
  4. Classify each use case against the relevant AI-specific risk framework and legal scope.

Explanation

Correct Answer

C:

Design human oversight with comprehension, time, access to evidence and power to override where required. The action addresses meaningful human oversight design rather than treating every AI control as interchangeable. The decisive signal to verify is that the reviewer's interface displays only a final score and an unchangeable approval button. This matters because a nominal human click is not the same as effective human control. Success should be measured by confirming that trained overseers can understand limits, investigate anomalies and exercise real intervention authority.

Incorrect Answers

A:

Assess whether the employment selection use falls within a high-risk AI category and plan controls under applicable timelines. A nominal human click is not the same as effective human control. Design human oversight with comprehension, time, access to evidence and power to override where required.

B:

Evaluate the applicable GPAI provider duties for technical and downstream documentation. A nominal human click is not the same as effective human control. Design human oversight with comprehension, time, access to evidence and power to override where required.

D:

Classify each use case against the relevant AI-specific risk framework and legal scope. A nominal human click is not the same as effective human control. Design human oversight with comprehension, time, access to evidence and power to override where required.

Artificial Intelligence Governance Professional (AIGP) Practice Tests