Security Controls and Monitoring AAISM Practice Questions (Page 3)

Questions 11–15

1. 11expert · hard

   A large enterprise has deployed an AI model for autonomous vehicle navigation. The model is critical for safety. The security team has detected unusual behavior that may indicate an adversarial attack. The team must decide on an incident response plan that balances the need for immediate action with the need for thorough investigation. What is the best approach?

   • AIgnore the behavior and continue monitoring for a few more days to gather more data.
   • BRetrain the model with new data and redeploy it immediately.
   • CActivate the incident response plan, isolate the affected model, and begin forensic analysis while maintaining a manual fallback.
   • DImmediately shut down the entire AI system and all related services.
   Reveal Solution Discussion Select an answer first
2. 12foundation · easy

   How can an organization demonstrate to an auditor that AI security controls are effective?

   • ABy providing evidence of regular security testing and monitoring results
   • BBy having the AI system certified by a vendor
   • CBy ensuring the AI system is not accessible from the internet
   • DBy stating that the controls are based on industry best practices
   Reveal Solution Discussion Select an answer first
4. 13expert · hard

   A company's AI system for predictive maintenance has been in production for a year. The system's performance has been stable, but the company is concerned about the risk of model drift due to changing equipment conditions. The company wants to implement a monitoring solution that detects drift early while minimizing false alarms. What is the best approach?

   • AImplement statistical process control (SPC) on the model's prediction errors and set control limits based on historical performance.
   • BMonitor the model's accuracy on a fixed test set and alert if accuracy drops below 90%.
   • CRetrain the model monthly to ensure it adapts to changing conditions.
   • DUse a simple threshold on the prediction error and alert if the error exceeds a certain value.
   Reveal Solution Discussion Select an answer first
5. 14application · medium

   A financial services firm is deploying an AI model for loan approval decisions. The firm must demonstrate to regulators that it has implemented appropriate security controls for the AI system. The security team has decided to adopt a widely recognized control framework. Which approach best aligns with the firm's regulatory requirement?

   • AUse a proprietary framework developed internally by the firm's data science team.
   • BApply the ISO/IEC 27001 standard directly to the AI model without any adaptation.
   • CAdopt the OWASP Top 10 for Large Language Model Applications as the sole control framework.
   • DImplement the NIST AI Risk Management Framework (AI RMF) and map its controls to the firm's existing regulatory obligations.
   Reveal Solution Discussion Select an answer first
6. 15foundation · easy

   What is the primary goal of an adversarial attack on an AI model?

   • ATo cause the model to make incorrect predictions
   • BTo corrupt the training data
   • CTo gain unauthorized access to the model's API
   • DTo steal the model's architecture
   Reveal Solution Discussion Select an answer first