Question 440
An Azure AI deployment requires centralized policy to block unsupported public endpoints. Which control design should be prioritized to meet the business requirement?
Explanation
Correct Answer
map AI resources to benchmark security controls and enforce private connectivity policies. This is the most relevant choice because AI workloads need the same enforceable cloud benchmarks as other sensitive services. An acceptance test should demonstrate that a newly created AI endpoint cannot enable a prohibited public exposure configuration.
Incorrect Answers
This measure is suited to a different threat model: A security team wants to detect harmful model behavior before onboarding a new AI workflow. Threat modeling for AI includes malicious prompts, tools, data, and outputs. It does not directly establish that a newly created AI endpoint cannot enable a prohibited public exposure configuration.
It addresses a neighboring control requirement, not this one: A developer connects a generative AI agent to production customer records with unrestricted tool calls. AI agents must respect explicit authorization outside natural-language instructions. The present scenario specifically needs a newly created AI endpoint cannot enable a prohibited public exposure configuration.
That recommendation concerns An enterprise stores confidential prompts in telemetry without retention controls. AI observability must not create a parallel sensitive-data exposure. Here the required proof is that a newly created AI endpoint cannot enable a prohibited public exposure configuration.