Question 860
After correcting this architecture issue: Employees use sanctioned and unknown generative AI websites, and security wants visibility across both. Which ongoing control belongs in the runbook?
Explanation
Correct Answer
Reassess discovered AI application inventory and sanctioned-use decisions monthly. This lowers the risk linked to: Traditional URL categorization may lack granular emerging generative-AI app identification. Verification should show: Reporting distinguishes recognized sanctioned AI apps from unauthorized new AI destinations.
Incorrect Answers
Require red-team reruns for material model or system-prompt updates. This safeguard addresses another situation: A new enterprise chatbot model has not been assessed for prompt injection before production launch. It is less directly relevant than the control for: Traditional URL categorization may lack granular emerging generative-AI app identification.
Link model version hashes to scan reports in release records. This safeguard addresses another situation: A model registry contains third-party weights but no one has screened them for unsafe behavior. It is less directly relevant than the control for: Traditional URL categorization may lack granular emerging generative-AI app identification.
Include prompt log locations in cardholder data environment review. This safeguard addresses another situation: A credit-card support bot receives raw PAN numbers and stores them in unrestricted logs. It is less directly relevant than the control for: Traditional URL categorization may lack granular emerging generative-AI app identification.