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 2: AI Risk Management
Q661 A security manager is deciding who handles risk when a third-party AI system causes harm in deployment. Which conclusion is MOST accurate?
- A. The deployer is liable only if it modified the vendor's model
- B. Liability depends only on the vendor's marketing claims
- C. Using a third-party system does not alleviate the deployer's responsibility or liability
- D. Liability transfers fully to the provider once a contract is signed and the deployer has paid its subscription fees
Answer: C
The manual states that even though a system is provided by a third party, the deployer's responsibility and liability are not removed. A contract does not transfer all risk.
Domain 2: AI Risk Management
Q662 An organization subscribes to an AI service and wants to clarify who does what. Which approach BEST reflects the AI shared responsibility model?
- A. Define responsibilities contractually, including SLAs, incident response and notification
- B. Leave responsibilities undefined until an incident occurs, then negotiate who pays based on the findings
- C. Assume the provider handles every control because it hosts the model
- D. Assume the subscriber handles every control because it owns the data
Answer: A
The manual says shared responsibility, similar to cloud, must be addressed contractually with clear SLAs, incident response activities and notification requirements. Assuming one party handles everything creates gaps.
Domain 2: AI Risk Management
Q663 Which responsibility belongs MOST directly to the AI subscriber under the shared responsibility model?
- A. Documenting the provider's internal model architecture
- B. Maintaining the provider's compliance certifications
- C. Auditing AI systems and penetration testing deployed controls
- D. Protecting the provider's model weights and inference-time artifacts
Answer: C
Subscriber responsibilities include audits and testing of privacy and security control configurations. Protecting weights and artifacts and holding certifications are provider duties.
Domain 2: AI Risk Management
Q664 Which responsibility is MOST appropriate for an AI provider rather than the subscriber?
- A. Protecting training data, model weights, logs and inference-time artifacts
- B. Training the subscriber's staff on its own incident response plan
- C. Tracking the provenance of the subscriber's own input data and the transformations applied before inference
- D. Deciding which employees may access the AI system
Answer: A
Provider data security duties include protecting training data, model weights, parameters, system logs and inference-time artifacts. Subscribers handle their own data governance, staff training and access management.
Domain 2: AI Risk Management
Q665 A subscriber wants to confirm that its provider follows relevant standards. Which evidence is MOST useful?
- A. Verbal assurances from the provider's sales team
- B. The number of customers the provider has onboarded this year
- C. Compliance and audit certifications such as SOC 2, ISO and CSA STAR
- D. The provider's marketing brochures describing its leading position
Answer: C
The manual's shared responsibility model lists compliance and audit certifications such as SOC 2, ISO and CSA STAR as provider evidence. Marketing and verbal claims are not verifiable assurance.
Domain 2: AI Risk Management
Q666 When assessing a potential AI provider, which supply chain activity should the subscriber perform FIRST?
- A. Review only the provider's price list and delivery times, then sign once procurement has approved the budget
- B. Vet provider security, financial stability and reputation, and add notification terms
- C. Rely on the provider's own assertion that it has no sub-vendors
- D. Accept the provider's standard terms without negotiation
Answer: B
Subscribers should assess the posture of the provider and its vendors, including financial stability and reputation, and include security, notification and privacy terms in contracts. Price alone or unquestioned terms are inadequate.
Domain 2: AI Risk Management
Q667 A subscriber wants users to know when they are dealing with AI. Which shared responsibility element does this BEST address?
- A. Configuration management
- B. Data provenance tracking
- C. Transparency and explainability
- D. Identity and access management
Answer: C
The manual says subscribers should ensure users are fully aware of any interactions with an AI system and choose explainable models. Access management, provenance and configuration management address different concerns.
Domain 2: AI Risk Management
Q668 A company plans to add a vendor's AI system to an aging core platform. Which action should it take FIRST?
- A. Assume the AI system will work with any platform without changes
- B. Check the provider's integration capabilities and whether AI makes sense
- C. Replace the legacy platform immediately before assessing any options
- D. Sign a multi-year contract with the vendor and evaluate integration with the legacy platform in a later phase
Answer: B
The manual advises working with stakeholders to decide if AI makes sense and checking the provider's integration capabilities, including whether a complete overhaul is needed. Committing first or assuming compatibility is risky.
Domain 2: AI Risk Management
Q669 During legacy integration planning, a leader notes that the data feeding the old system is inconsistent. What is the MOST appropriate response?
- A. Delete the inconsistent data without consulting its owners
- B. Allow the vendor to decide what data is acceptable, since it knows its own model better than the data owners
- C. Proceed with deployment because models can correct any data problem
- D. Check whether data quality and availability are sufficient or governance must improve first
Answer: D
The manual lists checking whether data quality or availability is sufficient or will require improved governance and management. Models do not fix poor data, and deleting data without owners causes new problems.
Domain 2: AI Risk Management
Q670 A legacy modernization project is expected to uncover hidden technical, privacy and security debt. Which planning approach is MOST consistent with the manual?
- A. Assume the provider will fund unexpected remediation
- B. Fix a rigid plan with a firm budget that cannot change during delivery, to prevent scope creep from legacy issues
- C. Skip estimating effort since modernization is always quick
- D. Use a flexible methodology and estimate effort, duration and funding needs
Answer: D
The manual advises a flexible methodology because modernizing uncovers technical, privacy and security debt, and an understanding of effort, duration and funds because upgrading legacy systems is not easy or inexpensive.