AI Risk Scenario Identification and Assessment (e.g., Threats, Vulnerabilities, and Attacks) AAIR Practice Questions (Page 5)

Questions 21–25

1. 21foundation · easy

   Which category of threat to an AI system is primarily concerned with an attacker stealing the model's architecture, parameters, or training data?

   • AModel theft
   • BData poisoning
   • CInference attack
   • DAdversarial attack
   Reveal Solution Discussion Select an answer first
2. 22application · medium

   A government agency uses an AI system to screen job applications. The system is publicly accessible and the agency has published detailed documentation about the model's architecture and training data. An activist group with moderate technical skills has expressed interest in disrupting the system. Which factor MOST increases the likelihood of a successful attack on this system?

   • AThe attacker's high level of intent
   • BThe system's high exposure due to public accessibility and detailed documentation
   • CThe attacker's advanced capability
   • DThe system's low impact on organizational objectives
   Reveal Solution Discussion Select an answer first
4. 23application · medium

   A financial institution uses a machine learning model to approve credit card transactions. The model was trained on historical transaction data that was collected from a public dataset. Recently, an attacker has been submitting fraudulent transactions that are designed to look like legitimate ones, causing the model to approve them. Which of the following is the MOST likely threat category for this attack?

   • AMembership inference
   • BModel theft
   • CModel evasion
   • DData poisoning
   Reveal Solution Discussion Select an answer first
5. 24foundation · easy

   Which of the following is a vulnerability in the deployment environment of an AI system?

   • AUse of a popular machine learning framework
   • BUnpatched operating system on the inference server
   • CLarge training dataset
   • DHigh model accuracy on validation data
   Reveal Solution Discussion Select an answer first
6. 25foundation · easy

   Which factor is most directly considered when estimating the likelihood of an AI risk scenario?

   • AThe potential financial impact
   • BThe number of stakeholders affected
   • CThreat actor capability and intent
   • DThe model's accuracy on the test set
   Reveal Solution Discussion Select an answer first