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
Q431 A company adopts a new AI architecture. Which risk should the security team MOST anticipate?
- A. All existing controls remain adequate because AI uses the same networks
- B. Security reviews become unnecessary once the vendor provides a datasheet
- C. AI removes the need to retest controls on the legacy platform
- D. Existing low-risk controls may become high risk under the new architecture
Answer: D
The manual warns that controls that were low risk may become high risk with new AI architecture. Existing controls must be reassessed. DevSecOps remains necessary with AI security considered throughout.
Domain 3: AI Technologies and Controls
Q432 How does the use of AI MOST affect DevSecOps and the secure development life cycle?
- A. DevSecOps is replaced entirely by a separate AI-only delivery process, because models cannot follow conventional pipelines
- B. Security testing is no longer needed because models are statistical
- C. The frameworks stay the same, but AI security must be considered throughout the architecture
- D. Only data scientists are responsible for security of the pipeline
Answer: C
DevSecOps and the SDLC do not specifically change, but the enterprise must ensure AI security is considered throughout, including infrastructure designed for secure AI. They are not replaced, and testing remains needed.
Domain 3: AI Technologies and Controls
Q433 Why is data dependency a MAJOR architectural concern for AI systems?
- A. Model behavior depends on data availability, quality and lineage across pipelines
- B. Data is needed only at deployment, so training pipelines are not a concern
- C. Data dependency applies only to relational databases used by the application
- D. AI systems do not use data until the moment they are deployed
Answer: A
AI models depend on data throughout training and inference, so availability, quality and lineage are architectural concerns. They use data at all stages, not just deployment.
Domain 3: AI Technologies and Controls
Q434 A team must pick between several candidate models for a regulated use case. Which factor is MOST important in model selection?
- A. The largest parameter count available among the candidate models
- B. Fit with regulatory, explainability and societal requirements
- C. The newest release date among the candidate models on the market
- D. The most popular model among practitioners on social media
Answer: B
Model selection must consider regulatory and societal impact and fit for purpose, including explainability. Parameter count, release date and popularity do not ensure compliance.
Domain 3: AI Technologies and Controls
Q435 Which practice is MOST important for AI configuration management?
- A. Version-controlling models, parameters and configurations with documented changes
- B. Storing settings in personal notebooks maintained by each data scientist so that experiments stay flexible and fast
- C. Avoiding change records so that experiments are not slowed down
- D. Allowing ad hoc edits in production when engineers find a quick fix
Answer: A
Configuration management tracks versions and changes of models and parameters so behavior can be reproduced and audited. Ad hoc edits or personal storage break traceability.
Domain 3: AI Technologies and Controls
Q436 A model update changes outputs in production without any record of the approval. Which control gap is MOST apparent?
- A. Strong version control
- B. Weak AI change management
- C. Too many backups
- D. Excessive encryption
Answer: B
AI change management should control and document updates to models and data with approval and testing. Unapproved changes show a gap. The other options describe strengths or unrelated matters.
Domain 3: AI Technologies and Controls
Q437 Which life cycle phase is MOST appropriate for defining success metrics, goals and an initial risk assessment?
- A. Retire and decommission
- B. Deploy
- C. Plan and design
- D. Operate and monitor
Answer: C
Planning and design articulates concept, objectives and assumptions, defines success metrics and conducts initial risk assessment including ethics and societal impact. Operating and retiring occur later.
Domain 3: AI Technologies and Controls
Q438 Why does the manual stress risk consideration PRIMARILY in the plan and design phase?
- A. Only regulators care about early design risk, not the organization itself
- B. Without it, organizations risk failure, unnecessary risk, bias and lack of accountability
- C. The phase does not affect later work, since risk is reassessed at launch
- D. Risk is irrelevant until deployment, when real users interact with the system and the first incidents become visible
Answer: B
Without risk consideration in planning and design, organizations face unsuccessful outcomes and risk from lack of ethical alignment, bias and unclear accountability. Early decisions affect every later phase.
Domain 3: AI Technologies and Controls
Q439 During data collection, what is the FIRST thing that must be obtained for data used to create a new AI solution?
- A. A competitor's approval for use of similar datasets
- B. A marketing slogan approved for the data collection campaign
- C. A vendor discount on bulk data purchases
- D. Consent for collected data
Answer: D
Consent should be obtained for all collected data used in creating a new AI solution, otherwise privacy laws may be broken and litigation risk increases. The other items are not requirements.
Domain 3: AI Technologies and Controls
Q440 A dataset's documentation records collection methods, sources, preprocessing steps and composition. What is this documentation BEST called?
- A. Data sharding
- B. Data masking
- C. Data retention
- D. Data provenance
Answer: D
Data provenance documents how the dataset was created, its sources, preprocessing and composition for transparency and accountability. Masking hides values, retention governs keeping data, and sharding partitions it.