Question 410
A user can call a model deployment through a key but cannot access the Foundry project portal. What does this indicate?
Explanation
Correct Answer
Model inference credentials and management-plane project access can authorize different operations, so success in one does not prove the other.
Incorrect Answers
Assign the application's managed identity the required search data-plane role. It addresses another case: A Foundry application is authenticated but cannot query an Azure AI Search index because the service denies its requests. For that case, the distinguishing justification or acceptance check is: Authentication to Foundry does not independently confer permission to query a separate protected search resource. The present case instead states: A user can call a model deployment through a key but cannot access the Foundry project portal. It calls for Check resource API authentication separately from the user's interactive project RBAC. Showing success for the other case would not establish: Model inference credentials and management-plane project access can authorize different operations, so success in one does not prove the other.
The alternative, Use separate least-privilege identities for deployment and secret administration, belongs to a different technical condition: A workload identity that deploys Foundry models also has permission to rotate production Key Vault secrets. There the source-specific mechanism or proof is: Deployment automation does not need broad secret-management authority; distinct roles reduce privilege escalation and accidental secret changes. The condition asked about here is: A user can call a model deployment through a key but cannot access the Foundry project portal. A suitable correction is Check resource API authentication separately from the user's interactive project RBAC; the targeted justification or result is: Model inference credentials and management-plane project access can authorize different operations, so success in one does not prove the other.
Revoke or rotate the compromised key and migrate the application to protected authentication. That operation would be relevant if: An AI application previously used a static API key that has been exposed in logs. Its corresponding reason or verification is: A credential leaked into logs must be invalidated; hiding the log does not remove the exposure already created. However the problem in this question is: A user can call a model deployment through a key but cannot access the Foundry project portal. The requirement instead calls for Check resource API authentication separately from the user's interactive project RBAC; the question-specific check is: Model inference credentials and management-plane project access can authorize different operations, so success in one does not prove the other.