Question 30
A network security design test investigates: An employee AI browser plugin is labeled approved but exports internal customer data to a new endpoint. Which verification demonstrates an effective solution?
Explanation
Correct Answer
A synthetic customer dataset upload is blocked even though the base application is sanctioned. It validates the recommended action: Classify the AI application by actual data flows and apply sanctioned-app DLP restrictions. The design principle is: Sanctioned status does not authorize every feature or destination of an AI application.
Incorrect Answers
Governance evidence maps major AI risks to accountable owners and measured mitigation tests. That observation supports a DIFFERENT situation: A company wants to satisfy NIST AI risk governance but measures only firewall block counts. Here the selected design is: Classify the AI application by actual data flows and apply sanctioned-app DLP restrictions.
Only approved AI tools meet the user-group and data-handling policy; unknown AI tools are monitored or blocked. That observation supports a DIFFERENT situation: A team allows all AI writing tools because their URLs appear in the productivity category. Here the selected design is: Classify the AI application by actual data flows and apply sanctioned-app DLP restrictions.
Only approved retrieval and tool endpoints receive traffic; other outbound connections fail. That observation supports a DIFFERENT situation: An internal RAG application permits unrestricted outbound tool requests to arbitrary web hosts. Here the selected design is: Classify the AI application by actual data flows and apply sanctioned-app DLP restrictions.