Question 230
An enterprise security architecture review finds: A business has approved generative AI assistants but does not know which unsanctioned web-based assistants employees actually use. Which evidence should be gathered first, before proposing a change?
Explanation
Correct Answer
For this case, start by compare AI Access Security app discovery capabilities and App-ID based browser telemetry. It establishes whether the reported condition—a business has approved generative AI assistants but does not know which unsanctioned web-based assistants employees actually use—has the assumed cause. The controlled remedy is then to use AI Access Security visibility and sanctioned/unsanctioned application controls for workforce usage, measured by whether the approved assistant is identifiable and unapproved AI destinations produce actionable policy logs.
Incorrect Answers
The alternative deals with “kubernetes-based AI services run in a shared cluster and compromise of one pod must not expose other model tenants”. This check distinguishes the real design constraint before the team commits to the change: apply network and workload microsegmentation around AI services plus appropriate runtime protection. The current question instead selects “Compare AI Access Security app discovery capabilities and App-ID based browser telemetry”.
This can be appropriate when a company deploys multiple AI inference endpoints and needs runtime visibility into malicious prompt attempts and responses, but it does not isolate the distinguishing requirement here: A business has approved generative AI assistants but does not know which unsanctioned web-based assistants employees actually use. The correct investigation decision is compare AI Access Security app discovery capabilities and App-ID based browser telemetry.
This choice targets the separate scenario “A customer wants to protect sensitive data entered into public AI tools while keeping certain sanctioned assistants accessible”. A later proof of effectiveness is that approved AI interactions work while a test restricted document upload produces a DLP event. The more relevant answer here is “Compare AI Access Security app discovery capabilities and App-ID based browser telemetry”.