This is not a replacement for the official study guide. Please read the official study guide first — use this exam prep only to reinforce or practice your understanding.
Free AAISM exam practice questions with answers and explanations, organized by domain and part.
Domain 1: AI Governance and Program Management
Q521 Recovery after an AI incident involving a poisoned model should MOST appropriately include which step?
- A. Skipping validation to restore service as quickly as possible
- B. Retraining on the same compromised data to save time
- C. Returning the poisoned model to service while monitoring it more closely for unusual behavior over several weeks
- D. Restoring a known good model version and validating it before returning to service
Answer: D
Recovery means restoring from a trusted model or retraining on clean data and validating before service resumes. Returning the poisoned model or reusing compromised data repeats the problem.
Domain 1: AI Governance and Program Management
Q522 Which use case of AI-driven incident response provides the MOST direct benefit in speeding containment?
- A. Automated triage and playbook execution for known incident patterns
- B. Reducing logging so that processing is faster
- C. Disabling alerts so analysts are not interrupted
- D. Eliminating all human review so that responses are instant and consistent across every incident type
Answer: A
Automating triage and playbooks accelerates routine response. Removing humans, disabling alerts or reducing logging weakens response.
Domain 1: AI Governance and Program Management
Q523 Which automation practice is MOST important for incident response automation?
- A. Never test automations because testing slows the response
- B. Automate every action with no review so the response is consistent, because human judgment adds delay and variation
- C. Hide the automation from the team to reduce resistance
- D. Keep humans in the loop for critical actions and regularly test and tune automations
Answer: D
Best practice keeps human oversight for critical steps and tests and tunes automations. Unreviewed, untested or hidden automations create unintended consequences.
Domain 1: AI Governance and Program Management
Q524 An analyst compares formats of dates across datasets and finds differing patterns. Which data quality dimension is MOST affected?
- A. Consistency
- B. Timeliness
- C. Completeness
- D. Validity
Answer: A
Consistency refers to uniform formats, lengths and metadata across datasets. Timeliness relates to currency, completeness to missing fields, and validity to adherence to business rules.
Domain 1: AI Governance and Program Management
Q525 A navigation system uses outdated coordinates to compute routes. Which data quality dimension is MOST affected?
- A. Consistency
- B. Validity
- C. Timeliness
- D. Accuracy of spelling only
Answer: C
Timeliness means data is up to date and available when needed, as with GPS coordinates. Consistency concerns uniform formats and validity concerns rules.
Domain 1: AI Governance and Program Management
Q526 Which data classification approach is MOST useful to concentrate data quality efforts on the data that matters most?
- A. Ignoring classification and relying on database naming conventions
- B. Tagging critical data elements and mapping them to a business glossary
- C. Tagging all data as public so that access is simpler to manage
- D. Deleting all unclassified data to reduce the volume to be reviewed
Answer: B
Critical data elements are tagged and tracked and mapped to an approved business glossary to make them manageable. Treating all data the same or deleting it does not focus effort.
Domain 1: AI Governance and Program Management
Q527 A national AI strategy emphasizes transparency, robustness, accountability and respect for human rights. Which source is MOST likely the origin of these principles?
- A. SOX
- B. GDPR Article 5 only
- C. OECD AI Principles
- D. PCI DSS
Answer: C
The OECD AI Principles promote inclusive growth, human rights, transparency, robustness, security, safety and accountability. PCI DSS and SOX address cards and financial reporting. GDPR principles are privacy focused.
Domain 1: AI Governance and Program Management
Q528 An AI strategy defines opportunities, vision and mission. Which activity BEST ensures that AI efforts support the organization's goals?
- A. Value alignment through frameworks, audits, human engagement and organizational change
- B. Selecting tools by popularity among competitors in the same sector, since peer adoption indicates fitness for purpose
- C. Avoiding stakeholder input to maintain consistent messaging
- D. Delegating the strategy to a vendor with experience in AI
Answer: A
Value alignment enablers include frameworks and guidance, human engagement, organizational change and audits. Popularity, delegation and avoiding input do not align AI with values.
Domain 1: AI Governance and Program Management
Q529 Which statement about value alignment is MOST accurate?
- A. It differs by organization and country and should consider the intended user base
- B. It is only about cost and has no connection to ethical principles, since value alignment is a purely financial measure
- C. It is unrelated to law and depends only on technical preference
- D. It is identical everywhere so one global standard can be applied
Answer: A
Alignment depends on the intended user base, culture, legal compliance and values, so it differs between organizations and countries. It is not uniform and relates to law and ethics.
Domain 3: AI Technologies and Controls
Q530 A fraud team needs a model whose individual decisions can be explained to regulators. Which model is MOST suitable?
- A. An ensemble of hundreds of boosted models whose combined output is averaged
- B. A decision tree or logistic regression whose decision paths can be traced and communicated
- C. A generative model that summarizes each case in natural language for reviewers
- D. A deep neural network with millions of parameters tuned for maximum accuracy
Answer: B
Simple models such as decision trees or logistic regression give traceable decision paths that regulators can follow. Deep networks and large ensembles are more accurate in some cases but harder to interpret, and a generative summary does not explain the decision itself.