IAPP

Free AIGP - AI Governance Professional Practice Questions

Test your knowledge with 10 free sample practice questions for the AIGP - AI Governance Professional certification. Each question includes a detailed explanation to help you learn.

10 Questions

No time limit

Free - No signup required

Disclaimer: These are original, AI-generated practice questions created by ProctorPulse for exam preparation purposes. They are not sourced from any official exam and are not affiliated with or endorsed by IAPP. Use them as a study aid alongside official preparation materials.

Question 1: What is the appropriate rollback decision given these circumstances?

A. Immediately roll back to manual review for all imaging sources
B. Roll back AI processing only for the affected scanner brand images (Correct Answer)
C. Continue AI processing with mandatory radiologist confirmation added
D. Maintain current operations and continue monitoring diagnostic impact

Explanation: A targeted rollback for the affected scanner brand is proportionate to the identified risk. The 12% inconsistency rate is significantly elevated and represents potential patient safety risk, warranting rollback even without confirmed diagnostic impact. A full rollback is unnecessarily disruptive since the system performs normally with other scanner brands. Monitoring without action ignores a known elevated risk.

Question 2: (Select all that apply) What incident management gap does this logging classification reveal?

A. The system lacks adequate pedestrian detection algorithm capability
B. Incident severity classification criteria are poorly calibrated (Correct Answer)
C. The vehicle should have engaged the emergency braking protocols (Correct Answer)
D. Near-miss events should not require any formal investigation steps

Explanation: A physical collision should never be classified as a 'resolved near-miss' regardless of outcome severity. This reveals poorly calibrated incident classification criteria in the system's logging framework. Accurate severity classification is critical for incident management because it determines escalation paths, investigation requirements, regulatory reporting obligations, and systemic risk assessment. Both B and C are correct.

Question 3: What governance control could have most effectively prevented this incident?

A. Pre-publication human review of all AI-generated content output
B. Automated plagiarism detection integrated into the output pipeline (Correct Answer)
C. Contractual indemnification clauses with the AI tool's provider
D. Training staff to recognize common AI-generated text style cues

Explanation: Automated plagiarism and similarity detection integrated into the AI output pipeline is the most effective preventive control. Human review cannot reliably detect similarity to copyrighted works at scale, contractual clauses address liability but not prevention, and style recognition does not identify specific copyright infringement. An automated check provides systematic, scalable prevention.

Question 4: What concern should the organization raise about the proposed audit scope?

A. The audit period should be limited to the most recent six months
B. Model architecture details are proprietary and should be fully excluded
C. Candidate personal data in hiring records requires privacy safeguards (Correct Answer)
D. Training data access should be limited to statistical summaries only

Explanation: While comprehensive audits require broad access, providing two years of hiring decision records containing candidate personal data raises significant privacy concerns. The organization should ensure appropriate data protection safeguards—such as anonymization, access controls, and data processing agreements—before sharing personal data with auditors, balancing audit thoroughness with privacy obligations.

Question 5: What governance action is most appropriate regarding this system?

A. Continue operations since performance benchmarks are still being met
B. Reduce the system's weight in parole decisions to advisory role only
C. Commission a new independent validation study to reassess accuracy
D. Initiate decommissioning and transition to a validated methodology (Correct Answer)

Explanation: When the scientific foundation of an AI system is discredited, continued use is inappropriate regardless of historical performance metrics. Performance benchmarks measure consistency with past behavior, not validity. A system based on discredited methodology should be decommissioned and replaced, as it produces systematically invalid outputs even if those outputs are internally consistent.

Question 6: What post-market surveillance action should the vendor take?

A. Issue a targeted software update only for the reporting hospital
B. Notify all deploying hospitals and assess cross-site performance (Correct Answer)
C. Investigate the specific hospital's data quality and integration
D. Add a warning label for this condition to product documentation

Explanation: Post-market surveillance obligations require vendors to treat a performance issue identified at one deployment site as potentially systemic. Notifying all hospitals and conducting cross-site assessment ensures that other deployments are not silently affected. A targeted fix or documentation update would be insufficient without first understanding whether the issue exists across all deployments.

Question 7: (Select all that apply) What is the most likely explanation for this production discrepancy?

A. The tool's bias emerged from interaction with live applicant data (Correct Answer)
B. Pre-deployment testing used synthetic data that masked real-world bias (Correct Answer)
C. Candidates with disabilities self-selected out of the hiring process
D. The disability variable was deliberately excluded from model features

Explanation: When pre-deployment testing shows acceptable results but production reveals significant disparities, the most likely explanation is that real-world data distributions differ from test conditions. Live applicant data may contain patterns—such as resume formatting differences, employment gap correlations, or accommodation-related language—that the model exploits as proxies, creating bias that synthetic or curated test data could not surface. Both A and B are correct.

Question 8: What does this post-market surveillance finding most directly indicate?

A. The system requires immediate retraining with current research data
B. The original training methodology was fundamentally flawed overall
C. Data drift monitoring controls were insufficient or entirely absent (Correct Answer)
D. The system should be decommissioned and replaced with a new one

Explanation: A gradual decline from 82% to 61% over 18 months is a classic indicator of data drift that should have been detected much earlier by adequate monitoring controls. While retraining may be the remediation, the surveillance finding most directly reveals that continuous monitoring for distribution shift was either not implemented or not acting on early warning signals.

Question 9: What should the incident management team's first remediation step be?

A. Honor all fabricated offers in order to maintain customer trust
B. Add output validation filters to block any unauthorized promotions
C. Issue a public statement explaining the AI system's malfunction
D. Disable the chatbot and switch to human-only customer service (Correct Answer)

Explanation: The immediate remediation priority is to stop the source of harm by disabling the malfunctioning chatbot and reverting to human agents. Honoring fabricated offers creates financial liability without addressing the root cause. Output filters and public communications are appropriate follow-up steps, but the first action must stop the system from generating further unauthorized commitments.

Question 10: How should the audit scope be adjusted to maintain integrity within constraints?

A. Accept English-only scope and plan multilingual audits for later
B. Prioritize testing the top five languages ranked by user volume
C. Include a representative sample across all twelve languages served (Correct Answer)
D. Extend the budget to cover comprehensive testing in all languages

Explanation: Audit integrity requires representative coverage of all operational contexts. A stratified sample across all 12 languages, even if smaller per language, is methodologically superior to deep testing of only a subset. Content moderation risks can manifest uniquely in different languages, and excluding any language creates blind spots in the audit's findings.

Question 1Medium

What is the appropriate rollback decision given these circumstances?

AImmediately roll back to manual review for all imaging sources

BRoll back AI processing only for the affected scanner brand images

CContinue AI processing with mandatory radiologist confirmation added

DMaintain current operations and continue monitoring diagnostic impact

Question 2Medium

(Select all that apply) What incident management gap does this logging classification reveal?

(Select all that apply)

AThe system lacks adequate pedestrian detection algorithm capability

BIncident severity classification criteria are poorly calibrated

CThe vehicle should have engaged the emergency braking protocols

DNear-miss events should not require any formal investigation steps

Question 3Medium

What governance control could have most effectively prevented this incident?

APre-publication human review of all AI-generated content output

BAutomated plagiarism detection integrated into the output pipeline

CContractual indemnification clauses with the AI tool's provider

DTraining staff to recognize common AI-generated text style cues

Question 4Medium

What concern should the organization raise about the proposed audit scope?

AThe audit period should be limited to the most recent six months

BModel architecture details are proprietary and should be fully excluded

CCandidate personal data in hiring records requires privacy safeguards

DTraining data access should be limited to statistical summaries only

Question 5Hard

What governance action is most appropriate regarding this system?

AContinue operations since performance benchmarks are still being met

BReduce the system's weight in parole decisions to advisory role only

CCommission a new independent validation study to reassess accuracy

DInitiate decommissioning and transition to a validated methodology

Question 6Medium

What post-market surveillance action should the vendor take?

AIssue a targeted software update only for the reporting hospital

BNotify all deploying hospitals and assess cross-site performance

CInvestigate the specific hospital's data quality and integration

DAdd a warning label for this condition to product documentation

Question 7Hard

(Select all that apply) What is the most likely explanation for this production discrepancy?

(Select all that apply)

AThe tool's bias emerged from interaction with live applicant data

BPre-deployment testing used synthetic data that masked real-world bias

CCandidates with disabilities self-selected out of the hiring process

DThe disability variable was deliberately excluded from model features

Question 8Hard

What does this post-market surveillance finding most directly indicate?

AThe system requires immediate retraining with current research data

BThe original training methodology was fundamentally flawed overall

CData drift monitoring controls were insufficient or entirely absent

DThe system should be decommissioned and replaced with a new one

Question 9Easy

What should the incident management team's first remediation step be?

AHonor all fabricated offers in order to maintain customer trust

BAdd output validation filters to block any unauthorized promotions

CIssue a public statement explaining the AI system's malfunction

DDisable the chatbot and switch to human-only customer service

Question 10Hard

How should the audit scope be adjusted to maintain integrity within constraints?

AAccept English-only scope and plan multilingual audits for later

BPrioritize testing the top five languages ranked by user volume

CInclude a representative sample across all twelve languages served

DExtend the budget to cover comprehensive testing in all languages

Ready for More?

These 10 questions are just a preview. Create a free account to practice up to 3 topics with 50 questions per day — or upgrade to Pro for unlimited access.

Ready to Pass the AIGP - AI Governance Professional?

Join thousands of professionals preparing for their AIGP - AI Governance Professional certification with ProctorPulse. AI-generated questions, detailed explanations, and progress tracking.