Question 20
Before changing controls, a principal architect receives this report: A multinational enterprise uses different risk tolerances for internal models and publicly hosted AI applications. Which check is most relevant?
Explanation
Correct Answer
For this case, start by classify AI model ownership, data exposure, access identities and applicable frameworks. It establishes whether the reported condition—a multinational enterprise uses different risk tolerances for internal models and publicly hosted AI applications—has the assumed cause. The controlled remedy is then to define distinct sanctioned-app and sensitive-content controls for internal versus external AI endpoints, measured by whether controlled trials show appropriate differences while high-risk data remains protected in both.
Incorrect Answers
Applying this option would direct attention to a healthcare AI assistant must process selected clinical inputs, but organization policy forbids sending identifiable patient content to external models. In the reported design—a multinational enterprise uses different risk tolerances for internal models and publicly hosted AI applications—the needed investigation statement is classify AI model ownership, data exposure, access identities and applicable frameworks.
That alternative belongs to the case: Legal counsel asks whether a consumer AI integration satisfies GDPR and NIST risk management obligations without evidence of actual data flow. It is not sufficient evidence or remediation for a multinational enterprise uses different risk tolerances for internal models and publicly hosted AI applications. The case-specific choice is classify AI model ownership, data exposure, access identities and applicable frameworks.
This choice targets the separate scenario “security teams discover a new AI app used by staff without a completed risk assessment”. If supported by the evidence, the target design would classify the application and apply a temporary appropriately scoped policy until formal approval. The more relevant answer here is “Classify AI model ownership, data exposure, access identities and applicable frameworks”.