AI Models, Frameworks, Strategies, and Use Cases AAIR Practice Questions (Page 6)

Questions 26–30

1. 26foundation · easy

   Which governance touchpoint is most important during the monitoring stage of an AI model's lifecycle?

   • ATracking model performance drift and data drift over time
   • BSelecting the model architecture
   • CDefining the initial training dataset
   • DDecommissioning the model immediately after deployment
   Reveal Solution Discussion Select an answer first
2. 27application · medium

   A financial services firm is integrating AI risk governance into its enterprise risk management framework. They have adopted NIST AI RMF and want to ensure that AI model risks are consistently identified and managed across the organization. What is the most effective way to achieve this integration?

   • AMap AI-specific risks to existing enterprise risk categories and assign ownership to risk owners
   • BUse only the NIST AI RMF and disregard other enterprise risk frameworks
   • CCreate a separate AI risk register that is managed independently by the data science team
   • DHave the legal department review all AI models for compliance without involving risk management
   Reveal Solution Discussion Select an answer first
4. 28foundation · easy

   Which AI-specific risk is most concerned with a model's vulnerability to adversarial inputs that cause incorrect predictions?

   • AExplainability
   • BRobustness
   • CBias
   • DPrivacy
   Reveal Solution Discussion Select an answer first
5. 29expert · hard

   A bank has deployed a machine learning model for credit scoring. The model was validated using historical data, but after deployment, the bank notices that the model's performance is declining because of economic changes. The bank must decide how to respond while maintaining regulatory compliance. What is the most appropriate action?

   • APause the model's use and conduct a thorough review before making changes
   • BIncrease the model's complexity to better adapt to economic changes
   • CImmediately retrain the model with the most recent data and redeploy it
   • DAdjust the model's decision threshold to compensate for the performance decline
   Reveal Solution Discussion Select an answer first
6. 30foundation · easy

   When an organization decides whether to build a custom AI solution or buy a commercial AI product, which factor is most directly related to the organization's risk appetite?

   • AThe level of control and customization required versus the risk of vendor lock-in
   • BThe programming language used by the development team
   • CThe number of GPUs available in the data center
   • DThe availability of open-source libraries
   Reveal Solution Discussion Select an answer first