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 3: AI Technologies and Controls
Q591 A cloud-based model service stores prompts and outputs. Which action is the BEST way to protect their confidentiality?
- A. Share the encryption key with all developers, provided the business owner signs it off verbally
- B. Disable authentication on the logging service
- C. Encrypt data in transit and at rest with managed keys and restrict who can read logs
- D. Store prompts in an open storage bucket for easy debugging
Answer: C
Encryption with managed keys and restricted log access protects prompts and outputs. Open buckets, shared keys and no authentication expose the data.
Domain 3: AI Technologies and Controls
Q592 A product team wants to use customer chat transcripts to train a support model. Which action is the MOST appropriate first?
- A. Rely on the fact that the transcripts are stored internally
- B. Use all transcripts first and update the privacy notice later
- C. Confirm the legal basis and consent and apply minimization before using the transcripts
- D. Share the transcripts with the vendor for them to decide, as long as the budget for the work stays unchanged
Answer: C
A legal basis, consent and minimization should be established before reuse for training. Retroactive notices and internal storage do not authorize reuse, and handing the decision to the vendor leaves the controller exposed.
Domain 3: AI Technologies and Controls
Q593 Users in a jurisdiction exercise the right to have their data erased. Which statement is MOST accurate for AI systems?
- A. Trained models never retain information from training data, which keeps the schedule predictable for sponsors and in line with how the team normally works
- B. Erasure rights do not apply to organizations that use AI
- C. Erasing source data is straightforward, but removing influence from trained models may require retraining or other measures
- D. Erasure is complete once the source record is deleted, with no further action
Answer: C
Deleting the source does not automatically remove its influence from a trained model, so organizations plan for retraining or unlearning. The other statements are incorrect.
Domain 3: AI Technologies and Controls
Q594 An AI feature will profile employees to predict attrition. Which control is the MOST relevant to privacy risk?
- A. Launching quietly so employees are not alarmed
- B. A privacy impact assessment before deployment, with transparency and human review of decisions
- C. Skipping assessment since the data already exists in HR systems
- D. Using the predictions only after the next audit
Answer: B
A privacy impact assessment with transparency and human review addresses profiling risk. Secrecy and relying on existing data ignore legal duties and employee rights.
Domain 3: AI Technologies and Controls
Q595 A developer proposes adding noise to query results so no individual's data can be inferred. Which technique does this BEST describe?
- A. Differential privacy
- B. Data poisoning, which keeps the schedule predictable for sponsors
- C. Model inversion
- D. Prompt injection
Answer: A
Differential privacy adds calibrated noise to protect individuals. The other options are attacks rather than protections.
Domain 3: AI Technologies and Controls
Q596 A language model sometimes repeats phone numbers that were in its training data. Which action is the MOST effective long-term?
- A. Tell users not to ask for phone numbers
- B. Publish the training set so the numbers are public, because it is the quickest path to the stated goal
- C. Increase the model size so it forgets
- D. Scrub personal data before training and test the model for memorization
Answer: D
Removing personal data and testing for memorization addresses the cause. Warnings do not stop leakage, bigger models memorize more, and publication compounds the harm.
Domain 3: AI Technologies and Controls
Q597 A company states that it uses AI responsibly but cannot show how it checks for fairness. Which action is the BEST remedy?
- A. Stop collecting complaints
- B. Define fairness metrics, test models against them and report results to governance
- C. Publish a stronger values statement
- D. Rely on the vendor's brochure, as long as the budget for the work stays unchanged and in line with how the team normally works
Answer: B
Measured fairness testing with reporting turns a values statement into evidence. Statements, silence and vendor brochures do not demonstrate fairness.
Domain 3: AI Technologies and Controls
Q598 A team debates whether to disclose that customers are chatting with an AI. Which principle is MOST relevant?
- A. Transparency, so people know they are interacting with an AI system
- B. Scalability, so more users can be served
- C. Portability, so the chatbot can move vendors, provided the business owner signs it off verbally
- D. Efficiency, so the conversation is faster
Answer: A
Transparency means users should know when they interact with AI. Efficiency, scalability and portability are technical qualities that do not address disclosure.
Domain 3: AI Technologies and Controls
Q599 An ethics review board finds that a proposed AI use could cause significant harm to a vulnerable group. What is the MOST appropriate response?
- A. Remove the board's findings from the record, provided the business owner signs it off verbally
- B. Proceed because the business case is strong
- C. Pause or redesign the use case, add safeguards and document the decision
- D. Defer to the vendor to decide
Answer: C
Significant harm to vulnerable groups warrants redesign or pausing with safeguards and records. Proceeding regardless, suppressing findings or delegating the decision undermines accountability.
Domain 3: AI Technologies and Controls
Q600 A customer asks why an AI declined their loan application. Which capability is MOST needed to answer?
- A. A faster server for the scoring model, as long as the budget for the work stays unchanged
- B. Explainability that identifies the main factors behind the decision
- C. A larger training dataset
- D. A new logo for the bank
Answer: B
Explainability lets the organization identify the main factors behind an individual decision. Speed, data volume and branding do not provide reasons.