4.2 Define Secure AI and Its Importance in Protecting AI Systems from Malicious Attacks and Misuse. GENERATIVE-AI-LEADER Practice Questions (Page 6)

Questions 26–28

1. 26foundation · easy

   Which type of information would Security Command Center provide about an AI workload?

   • AThe source code of the AI application
   • BThe exact accuracy score of a deployed model
   • CA list of security findings such as misconfigured IAM policies or vulnerable libraries
   • DThe cost breakdown of running AI workloads
   Reveal Solution Discussion Select an answer first
2. 27application · medium

   A company is training a model on sensitive customer data. They are concerned about a malicious insider who has access to the training environment. The insider could potentially exfiltrate the data or tamper with the training process. Which measure is most effective in mitigating this risk?

   • AUse a VPN to connect to the training environment.
   • BEncrypt the training data at rest and in transit.
   • CDisable all logging to reduce the attack surface.
   • DImplement strict IAM controls with least privilege and enable audit logging on all training data access.
   Reveal Solution Discussion Select an answer first
3. 28application · medium

   A startup is building a new AI product and wants to integrate security from the very beginning of the design process. They are concerned about the cost and effort of retrofitting security later. Which approach aligns with the principles of Google's Secure AI Framework (SAIF)?

   • AIncorporate security reviews and threat modeling into each phase of the AI development lifecycle, from data collection to deployment.
   • BRely on the cloud provider's infrastructure security to protect the AI application without additional measures.
   • CDefer security considerations until after the model is trained and ready for production.
   • DUse a single security tool that covers all aspects of the AI lifecycle.
   Reveal Solution Discussion Select an answer first