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
Q261 A team plans to train a model on generated content without review. What is the PRIMARY risk?
- A. Reduced network bandwidth during large training data transfers
- B. Subtle inaccuracies or bias propagating into the new model
- C. Excess encryption overhead slowing every training iteration
- D. Lower regulatory scrutiny because the data is synthetic
Answer: B
Generated results may be inaccurate and, when used for training, can create biased or nonfunctional models, so human review is needed. Bandwidth and encryption are unrelated. Review reduces the risk.
Domain 1: AI Governance and Program Management
Q262 A user's vague prompt yields an unexpected result. Which mitigation is MOST appropriate?
- A. Retire the model and replace it with a newer release
- B. Disable all prompt logging so earlier inputs cannot influence output
- C. Write precise, contextually relevant prompts with clear instructions
- D. Increase the model token limit so longer prompts are accepted
Answer: C
Output is heavily influenced by user input, so precise prompts reduce unexpected results. Disabling logging worsens oversight, and a larger limit does not clarify intent. Retiring the model is excessive.
Domain 1: AI Governance and Program Management
Q263 Which content is the MOST complete minimum for a documented AI business case?
- A. The vendor's price list, discount schedule, support terms and contract renewal dates
- B. A summary of competitor AI failures and lessons learned across the industry
- C. The problem, proposed solution, expected ROI, identified risk and an implementation plan
- D. Technical architecture diagrams, network topology and server sizing specifications
Answer: C
The manual lists the problem, how the solution solves it, ROI expectations, risk and mitigations, and a detailed implementation plan. Price or diagrams alone cannot justify the investment.
Domain 1: AI Governance and Program Management
Q264 An organization considers a widely used cloud AI service versus a brand-new model concept. Why is the established service the BEST initial business case?
- A. It is always cheaper than any alternative in total cost of ownership
- B. Its use case, costs and benefits are well known, reducing variability
- C. It needs no governance oversight since the provider manages everything
- D. It carries no residual risk because many organizations already use it
Answer: B
Established solutions have known use cases, costs and benefits, which removes variability. Neither option is risk-free or free of governance. Novel concepts bring expectations without a clear path.
Domain 1: AI Governance and Program Management
Q265 Under which cost category does employee training for an AI solution BEST fall?
- A. Opportunity costs
- B. Implementation costs
- C. Operational costs
- D. Risk and compliance costs
Answer: B
Implementation costs include development, licensing and training for employees. Operational costs cover maintenance, energy and new salaries. Risk costs cover privacy protection and reputation management.
Domain 1: AI Governance and Program Management
Q266 Which cost would BEST be classified as a risk and compliance cost in an AI cost-benefit analysis?
- A. Energy consumption
- B. Developer salaries
- C. Licensing fees
- D. Reputation management
Answer: D
Risk and compliance costs include privacy protection and reputation management. Energy is operational and licensing is implementation. Salaries for new AI staff are operational.
Domain 1: AI Governance and Program Management
Q267 A company with no in-house AI expertise needs a fast solution. Which approach is MOST reasonable?
- A. Postpone any AI initiative until internal staff can train a foundation model and operate the full stack themselves
- B. Engage a vendor or an off-the-shelf solution while defining shared responsibilities
- C. Let business units choose their own unvetted AI tools without central review
- D. Build a custom foundation model from scratch using contract developers alone
Answer: B
Without in-house expertise, engaging a vendor or collaborating through groups such as CoSAI may be best, with responsibilities defined. Building from scratch ignores the skills gap. Unvetted tools add risk.
Domain 1: AI Governance and Program Management
Q268 Which is the MAIN benefit of hosting an AI solution internally rather than in the cloud?
- A. Greater control and ease of securing data and privacy
- B. Lower upfront implementation cost compared with a cloud subscription
- C. Easier scaling of resources up or down as demand changes
- D. Reduced operational overhead for system maintenance and upgrades
Answer: A
Internal hosting provides more control, customization and easier data security and privacy control. Scaling, lower upfront cost and less overhead are cloud benefits. Internal hosting needs more resources.
Domain 1: AI Governance and Program Management
Q269 Which limitation is MOST associated with a cloud-hosted AI solution?
- A. Inability to scale resources when demand rises sharply
- B. A need for additional on-site staff with specialist skills
- C. Vendor lock-in and less infrastructure control
- D. High capital outlay for computing and power resources
Answer: C
Cloud solutions bring less control and possible vendor lock-in plus data leaving the controlled environment. Scalability is a strength, and high capital outlay and staff needs are internal hosting limits.
Domain 1: AI Governance and Program Management
Q270 Training data subject to privacy law will be processed by a cloud AI provider. Which factor becomes MOST important?
- A. The number of model parameters in the hosted model
- B. The physical location of the hosting provider
- C. The vendor's published list of supported programming languages
- D. The provider's brand recognition among its competitors
Answer: B
When data is covered by privacy laws, the provider's physical location is of increased concern for compliance. Logos or laptop brands are irrelevant. Parameter count has no bearing on legal jurisdiction.