Threats and Vulnerabilities Specific to AI AAIA Practice Questions (Page 1)

Questions 1–5

1. 1expert · hard

   A hospital deploys a diagnostic AI model. The model was trained on data from a specific region and now shows declining accuracy as patient demographics change. The hospital wants to maintain accuracy without retraining from scratch. Which approach is most appropriate?

   • AModel inversion
   • BContinuous monitoring and periodic fine-tuning
   • CAdversarial training
   • DInput validation
   Reveal Solution Discussion Select an answer first
2. 2application · medium

   A financial firm deploys a credit scoring model. The model is highly accurate but its decisions are opaque, making it difficult for auditors to understand why a loan was denied. This lack of transparency creates a compliance risk. Which AI-specific deployment vulnerability is this?

   • AData poisoning
   • BLack of explainability
   • CPrompt injection
   • DModel drift
   Reveal Solution Discussion Select an answer first
4. 3application · medium

   A healthcare startup trains a diagnostic model on a public dataset of medical images. After deployment, the model performs well on most patients but consistently misdiagnoses a rare condition. An audit reveals that a small number of images in the training set were deliberately mislabeled to cause this specific failure. Which AI-specific threat occurred?

   • AModel extraction
   • BData poisoning
   • CPrompt injection
   • DAdversarial attack
   Reveal Solution Discussion Select an answer first
5. 4foundation · easy

   What is model drift in the context of AI deployment?

   • AThe model is moved to a different cloud provider
   • BThe model's performance degrades over time due to changes in the underlying data distribution
   • CThe model's architecture is modified during deployment
   • DThe model's training data is updated frequently
   Reveal Solution Discussion Select an answer first
6. 5foundation · easy

   A security auditor is reviewing the threat landscape for an AI-based fraud detection system. Which threat category is MOST directly associated with the AI system itself?

   • ADistributed denial-of-service (DDoS) attack on the network infrastructure
   • BData poisoning of the training dataset
   • CInsider threat from a disgruntled employee with administrative access
   • DSQL injection in the application database
   Reveal Solution Discussion Select an answer first