4.2 Define Secure AI and Its Importance in Protecting AI Systems from Malicious Attacks and Misuse. GENERATIVE-AI-LEADER Practice Questions (Page 4)

Questions 16–20

1. 16application · medium

   A company has multiple teams working on different AI projects in the same Google Cloud project. The security team wants to ensure that each team can only access the resources they need for their specific project. They also want to detect any unauthorized access attempts. Which combination of IAM and monitoring should they use?

   • ACreate custom IAM roles with least privilege for each team, and enable audit logging to track access attempts.
   • BGrant all teams the Owner role on the project to simplify management.
   • CDisable audit logging to reduce the amount of data stored.
   • DUse a single service account for all teams to simplify key management.
   Reveal Solution Discussion Select an answer first
2. 17foundation · easy

   In Google Cloud IAM, what is the recommended practice for granting permissions to AI resources?

   • AUse only basic roles because they are easier to remember
   • BGrant all users the Owner role to simplify management
   • CGrant the minimum permissions required for each user or service account
   • DAvoid using service accounts because they are less secure than user accounts
   Reveal Solution Discussion Select an answer first
4. 18foundation · easy

   During the deployment stage of an ML lifecycle, which security concern is most directly associated with exposing a model as an API endpoint?

   • APoisoning of the training dataset
   • BOverfitting to the validation set
   • CCorruption of the source code repository
   • DUnauthorized access to the model's predictions
   Reveal Solution Discussion Select an answer first
5. 19application · medium

   A company has deployed several gen AI models in production. The security team wants a centralized view of misconfigurations, vulnerabilities, and potential threats across all AI assets. They also want to detect when a model is being probed for adversarial inputs. Which Google Cloud tool should they use?

   • ACloud Logging
   • BSecurity Command Center
   • CCloud Monitoring
   • DCloud Armor
   Reveal Solution Discussion Select an answer first
6. 20foundation · easy

   Which benefit does Google's Secure AI Framework (SAIF) provide to an organization adopting generative AI?

   • AIt offers a consistent set of security practices that can be applied across different AI projects
   • BIt ensures that AI models never produce biased outputs
   • CIt automatically patches all vulnerabilities in open-source AI libraries
   • DIt eliminates the need for any additional security controls in the cloud environment
   Reveal Solution Discussion Select an answer first