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
Q671 An AI provider's model was trained on web content that included copyrighted works. Which risk should the customer be MOST concerned about?
- A. Unoriginal output may expose the customer to litigation and penalties
- B. The model will run slower because of larger input files
- C. The provider's invoices will increase automatically each month
- D. The model will refuse to answer questions about copyrighted topics
Answer: A
The manual warns that training data can include copyrighted material, producing output that is not truly original and exposing the organization to lawsuits and penalties. Speed and billing are unrelated.
Domain 2: AI Risk Management
Q672 A firm uses an AI model to generate code, and the model draws on the vendor's dataset. What should the firm do FIRST to reduce the IP ownership uncertainty?
- A. Assume all generated code belongs to the firm without written terms
- B. Rely on the vendor's assurance in an email from a sales representative
- C. Keep the generated code secret and avoid mentioning it to the vendor
- D. Put ownership terms in the contract and consult general counsel
Answer: D
The manual notes the true IP holder is uncertain when a model pulls from the provider's dataset, and says ownership provisions should be in contracts with legal consulted. Assumptions and informal assurances are insufficient.
Domain 2: AI Risk Management
Q673 A legal team asks why IP risk with an AI provider is more than a hypothetical concern. Which answer is MOST accurate?
- A. IP risk exists only if the model is open source
- B. IP risk only affects the provider and never the customer, because the provider holds the data licenses and the model
- C. Unclear ownership and sourcing can cost the enterprise control of valuable assets
- D. IP risk disappears once the output is reviewed by a manager
Answer: C
The manual says IP risk can lead to real financial, legal and operational consequences, and that without clarity on ownership, sourcing and compliance the enterprise may lose control of its valuable assets.
Domain 2: AI Risk Management
Q674 Which factor should be confirmed FIRST with an AI provider before integrating with a legacy environment?
- A. That the provider will remove the legacy system after integration
- B. That the provider will own all data processed by the system
- C. That the provider can postpone its own security testing until after the enterprise has completed integration
- D. That the provider can support the enterprise through the whole integration
Answer: D
The manual lists making sure the provider can work with the enterprise to deliver proper support throughout implementation and integration. It does not suggest provider ownership of data or removing the legacy system.
Domain 2: AI Risk Management
Q675 A COO asks how adopting AI will affect people who rely on the existing process. Which step from the manual answers this MOST directly?
- A. Limit communication to the technical team during integration
- B. Understand how AI will affect everyone in the existing process
- C. Remove the legacy process owners from the project team
- D. Delay mapping information flows until after go-live
Answer: B
The manual says to understand how AI will affect everyone involved in the existing process and information flows. Excluding owners or delaying mapping increases integration risk.
Domain 2: AI Risk Management
Q676 Generative AI systems learn from what users provide. Which concern does this raise MOST directly for IP and confidentiality?
- A. User input is always discarded immediately and never stored
- B. The system must obtain consent from each user before generating any output
- C. Only the provider's own data can affect model behavior
- D. User input may be retained by the system and later surface in outputs
Answer: D
The manual notes GenAI systems do not just generate content but learn from what users provide and share. This creates exposure for IP and confidential data unless contracts and usage rules address it.
Domain 2: AI Risk Management
Q677 A company manages software supply chain risk by checking each library individually. Which additional consideration does the manual stress MOST for AI?
- A. Ignore models and datasets because only the code libraries carry vulnerabilities that attackers can exploit
- B. Rely on the provider's assurance that no vulnerabilities exist
- C. Understand the entire AI supply chain, not only individual components
- D. Review the supply chain only after an incident
Answer: C
The manual says traditional supply chain risk addresses individual components, whereas AI supply chain management must consider entire systems and protect brand, reputation and trust. Waiting for an incident or trusting assurances is weak.
Domain 2: AI Risk Management
Q678 Which outcome is the PRIMARY reason the manual gives for managing AI software supply chain risk beyond reducing the attack surface?
- A. To protect the enterprise's brand, reputation and trust
- B. To reduce the cost of cloud storage
- C. To eliminate the need for vendor contracts
- D. To remove the need for monitoring models in production
Answer: A
The manual says managing AI software supply chain risk is needed not just to reduce attack surface or remediate vulnerabilities but to protect brand, reputation and trust. It does not remove contracts or monitoring.
Domain 3: AI Technologies and Controls
Q679 A hospital uses an AI model to suggest diagnoses. Which control BEST ensures safety for final decisions?
- A. Human in the loop, where a clinician approves the decision before action
- B. Logging disabled, so no record of the suggestion is created
- C. Periodic review of aggregate accuracy only, with no case-level review
- D. Full automation, where the model's output is applied without review so patients receive results immediately
Answer: A
HITL puts a person in the final decision-making role and requires explicit approval before the workflow continues, such as a doctor reviewing a diagnosis. Full automation and missing logs remove oversight.
Domain 3: AI Technologies and Controls
Q680 An organization adds a human in the loop to a high-volume AI workflow. Which consequence should it plan for FIRST?
- A. The model will no longer need monitoring or logging
- B. The workflow will become faster because humans decide quicker than models once the review step is added
- C. The human reviewer will no longer need training or guidance
- D. The workflow will run more slowly, so apply the control to specified conditions
Answer: D
The manual warns that HITL slows the speed of the AI system, so organizations should design carefully and require approval for specified conditions. It does not replace monitoring or training.