Question 560 A new chat integration publishes raw agent exceptions and request headers in the user interface when model serving temporarily fails. Which design change directly satisfies the stated constraints? Package and deploy the custom model through supported Model Serving rather than assuming a managed Foundation Model API serves arbitrary uploaded weights. Use an approved user-authorization pattern or enforce equivalent per-user data access checks behind the app service. Deploy a user-facing interface through a supported Databricks Apps or approved channel integration with authenticated backend agent calls. Translate backend failures into safe user-facing error states, redact internals and keep detailed diagnostics in protected logs. ExplanationCorrect Answer D:The action targets the stated cause: User interface error handling is a security boundary; propagating raw exceptions can expose credentials and system internals. Check the repair by: trigger a simulated endpoint outage and verify the UI shows a neutral retry path while protected logs retain trace IDs. Other Answers A:User interface error handling is a security boundary; propagating raw exceptions can expose credentials and system internals. B:User interface error handling is a security boundary; propagating raw exceptions can expose credentials and system internals. C:User interface error handling is a security boundary; propagating raw exceptions can expose credentials and system internals.
Question 561 An insurance chatbot should show policy coverage but not the claimant's national ID number present in a retrieved claim document. Which implementation decision best restores the intended behavior? Treat webpage text as untrusted evidence and enforce fixed trusted instructions and tool permissions regardless of embedded commands. Enforce tenant and row-level privileges before search and apply masking only as an additional output safeguard. Redact the national ID before model context construction and enforce user-level access controls on the underlying claim record. Apply sensitive-field minimization and redaction at trace ingestion as well as the output boundary. ExplanationCorrect Answer C:The action targets the stated cause: Masking reduces exposure in generated text, while record authorization is a separate prerequisite for retrieving sensitive data. Check the repair by: test a synthetic claim and verify neither model trace nor output contains the ID and unauthorized callers cannot fetch the record. Other Answers A:Masking reduces exposure in generated text, while record authorization is a separate prerequisite for retrieving sensitive data. B:Masking reduces exposure in generated text, while record authorization is a separate prerequisite for retrieving sensitive data. D:Masking reduces exposure in generated text, while record authorization is a separate prerequisite for retrieving sensitive data.
Question 562 An insurance chatbot should show policy coverage but not the claimant's national ID number present in a retrieved claim document. Which underlying technical principle explains why this occurs? Prompt injection exploits confusion between retrieved content and application instructions; source text cannot escalate its own authority. Redacted identifiers do not remove confidential amounts or metadata leaked across tenant boundaries. Masking reduces exposure in generated text, while record authorization is a separate prerequisite for retrieving sensitive data. Output filtering occurs after model processing and cannot prevent earlier diagnostic logs from collecting protected values. ExplanationCorrect Answer C:Redact the national ID before model context construction and enforce user-level access controls on the underlying claim record. Other Answers A:Masking reduces exposure in generated text, while record authorization is a separate prerequisite for retrieving sensitive data. B:That mechanism describes another problem: A banking retrieval index returns transaction totals for every tenant, and the application proposes replacing account numbers with asterisks after retrieval. It does not explain this case, where masking reduces exposure in generated text, while record authorization is a separate prerequisite for retrieving sensitive data. D:Masking reduces exposure in generated text, while record authorization is a separate prerequisite for retrieving sensitive data.
Question 563 An insurance chatbot should show policy coverage but not the claimant's national ID number present in a retrieved claim document. After correction, what targeted test should demonstrate success? Inject the malicious line and verify the agent neither reveals trusted configuration nor executes additional tools. Query as two tenants and show that each can retrieve only its own authorized transaction source IDs. Test a synthetic claim and verify neither model trace nor output contains the ID and unauthorized callers cannot fetch the record. Inspect raw and exported traces from a synthetic session and confirm regulated fields are absent. ExplanationCorrect Answer C:Masking reduces exposure in generated text, while record authorization is a separate prerequisite for retrieving sensitive data. Other Answers A:Test a synthetic claim and verify neither model trace nor output contains the ID and unauthorized callers cannot fetch the record. B:Test a synthetic claim and verify neither model trace nor output contains the ID and unauthorized callers cannot fetch the record. D:Test a synthetic claim and verify neither model trace nor output contains the ID and unauthorized callers cannot fetch the record.
Question 564 An insurance chatbot should show policy coverage but not the claimant's national ID number present in a retrieved claim document. Which release safeguard most directly addresses this failure class? Maintain defense-in-depth with pre-retrieval entitlements plus output redaction regression tests. Define protected-field classification, preprompt redaction and independent document-level authorization checks. Enforce logging schemas and automated sensitive-data scans on tracing pipelines. Maintain adversarial retrieval fixtures and keep execution policy outside the model's natural-language output. ExplanationCorrect Answer B:Masking reduces exposure in generated text, while record authorization is a separate prerequisite for retrieving sensitive data. Other Answers A:This control addresses a different exposure: A banking retrieval index returns transaction totals for every tenant, and the application proposes replacing account numbers with asterisks after retrieval. It would not prevent this failure, because masking reduces exposure in generated text, while record authorization is a separate prerequisite for retrieving sensitive data. C:That masking reduces exposure in generated text, while record authorization is a separate prerequisite for retrieving sensitive data. D:Masking reduces exposure in generated text, while record authorization is a separate prerequisite for retrieving sensitive data.
Question 565 A developer masks customer names in the final response but an internal MLflow trace still stores full names and addresses. Which design change directly satisfies the stated constraints? Validate all output fields for sensitive-data exfiltration including encoded variants, and prevent unauthorized source access before generation. Enforce tenant and row-level privileges before search and apply masking only as an additional output safeguard. Apply sensitive-field minimization and redaction at trace ingestion as well as the output boundary. Redact the national ID before model context construction and enforce user-level access controls on the underlying claim record. ExplanationCorrect Answer C:The action targets the stated cause: Output filtering occurs after model processing and cannot prevent earlier diagnostic logs from collecting protected values. Check the repair by: inspect raw and exported traces from a synthetic session and confirm regulated fields are absent. Other Answers A:Output filtering occurs after model processing and cannot prevent earlier diagnostic logs from collecting protected values. B:Output filtering occurs after model processing and cannot prevent earlier diagnostic logs from collecting protected values. D:Output filtering occurs after model processing and cannot prevent earlier diagnostic logs from collecting protected values.
Question 566 A developer masks customer names in the final response but an internal MLflow trace still stores full names and addresses. Which causal explanation best accounts for the failure? Output filtering occurs after model processing and cannot prevent earlier diagnostic logs from collecting protected values. Redacted identifiers do not remove confidential amounts or metadata leaked across tenant boundaries. Masking reduces exposure in generated text, while record authorization is a separate prerequisite for retrieving sensitive data. Encoding changes representation, not disclosure risk; a citation field can be an exfiltration channel just like prose. ExplanationCorrect Answer A:Apply sensitive-field minimization and redaction at trace ingestion as well as the output boundary. Other Answers B:That mechanism describes another problem: A banking retrieval index returns transaction totals for every tenant, and the application proposes replacing account numbers with asterisks after retrieval. It does not explain this case, where output filtering occurs after model processing and cannot prevent earlier diagnostic logs from collecting protected values. C:Output filtering occurs after model processing and cannot prevent earlier diagnostic logs from collecting protected values. D:Output filtering occurs after model processing and cannot prevent earlier diagnostic logs from collecting protected values.
Question 567 A developer masks customer names in the final response but an internal MLflow trace still stores full names and addresses. Which concrete acceptance test best demonstrates the proposed change works? Run encoded-data attack fixtures and inspect all tool arguments and returned fields for blocked private values. Inspect raw and exported traces from a synthetic session and confirm regulated fields are absent. Query as two tenants and show that each can retrieve only its own authorized transaction source IDs. Test a synthetic claim and verify neither model trace nor output contains the ID and unauthorized callers cannot fetch the record. ExplanationCorrect Answer B:Output filtering occurs after model processing and cannot prevent earlier diagnostic logs from collecting protected values. Other Answers A:Inspect raw and exported traces from a synthetic session and confirm regulated fields are absent. C:Inspect raw and exported traces from a synthetic session and confirm regulated fields are absent. D:Inspect raw and exported traces from a synthetic session and confirm regulated fields are absent.
Question 568 A developer masks customer names in the final response but an internal MLflow trace still stores full names and addresses. Which operational practice should be made a durable requirement? Maintain defense-in-depth with pre-retrieval entitlements plus output redaction regression tests. Enforce logging schemas and automated sensitive-data scans on tracing pipelines. Combine preprompt authorization, payload inspection and adversarial encoding tests. Define protected-field classification, preprompt redaction and independent document-level authorization checks. ExplanationCorrect Answer B:Output filtering occurs after model processing and cannot prevent earlier diagnostic logs from collecting protected values. Other Answers A:That output filtering occurs after model processing and cannot prevent earlier diagnostic logs from collecting protected values. C:Output filtering occurs after model processing and cannot prevent earlier diagnostic logs from collecting protected values. D:This control addresses a different exposure: An insurance chatbot should show policy coverage but not the claimant's national ID number present in a retrieved claim document. It would not prevent this failure, because output filtering occurs after model processing and cannot prevent earlier diagnostic logs from collecting protected values.
Question 569 A banking retrieval index returns transaction totals for every tenant, and the application proposes replacing account numbers with asterisks after retrieval. What should the implementation team change first to resolve the described defect? Verify contractual confidentiality and authorization before reusing the corpus; keep restricted contracts out of public retrieval scope. Enforce tenant and row-level privileges before search and apply masking only as an additional output safeguard. Redact the national ID before model context construction and enforce user-level access controls on the underlying claim record. Apply sensitive-field minimization and redaction at trace ingestion as well as the output boundary. ExplanationCorrect Answer B:The action targets the stated cause: Redacted identifiers do not remove confidential amounts or metadata leaked across tenant boundaries. Check the repair by: query as two tenants and show that each can retrieve only its own authorized transaction source IDs. Other Answers A:Redacted identifiers do not remove confidential amounts or metadata leaked across tenant boundaries. C:Redacted identifiers do not remove confidential amounts or metadata leaked across tenant boundaries. D:Redacted identifiers do not remove confidential amounts or metadata leaked across tenant boundaries.
Question 570 A banking retrieval index returns transaction totals for every tenant, and the application proposes replacing account numbers with asterisks after retrieval. Why does the current approach fail under these conditions? Assume replacing customer account numbers with asterisks automatically changes the underlying tenant permissions on search results. Output filtering occurs after model processing and cannot prevent earlier diagnostic logs from collecting protected values. Permission to process a document internally does not automatically grant public redistribution rights or confidentiality waivers. Redacted identifiers do not remove confidential amounts or metadata leaked across tenant boundaries. ExplanationCorrect Answer D:Enforce tenant and row-level privileges before search and apply masking only as an additional output safeguard. Other Answers A:A display mask does not implement tenant authorization. Different tenants can still receive confidential totals or metadata unless source access is restricted. B:Redacted identifiers do not remove confidential amounts or metadata leaked across tenant boundaries. C:That mechanism describes another problem: An agent indexes internal vendor contracts for enterprise employees and then exposes the same retrieval index to public customers. It does not explain this case, where redacted identifiers do not remove confidential amounts or metadata leaked across tenant boundaries.
