Question 440A policy agent reads a support article containing text that tells it to ignore its developer rules and reveal internal notes. Which preventive standard should the team adopt? Maintain a documented injection-regression suite and forbid unreviewed elevation of retrieved content into trusted instructions. Require pull-request review and change logs for system prompt modifications in production applications. Enforce prompt repository secret scanning and fixture anonymization before review and deployment. Test private connectivity and DNS changes before disabling public network paths in production. ExplanationCorrect Answer A:Maintain a documented injection-regression suite and forbid unreviewed elevation of retrieved content into trusted instructions. This reduces the failure risk because Retrieved documents can contain adversarial instructions; instruction hierarchy and explicit context boundaries reduce the chance of treating data as directives. Immediate action: Keep untrusted retrieved text and user content distinct from system instructions and test prompt-injection attempts during prompt design. Other Answers B:Prompt text is application behavior; version control enables traceable changes, reviews, and exact re-deployment of prior versions. The preventive control that addresses this condition is: Maintain a documented injection-regression suite and forbid unreviewed elevation of retrieved content into trusted instructions. C:Prompt source control should contain reusable instructions, not private customer inputs or API credentials that can leak through repository history. The preventive control that addresses this condition is: Maintain a documented injection-regression suite and forbid unreviewed elevation of retrieved content into trusted instructions. D:Private endpoints require consistent name resolution and routing for approved resources; blocking public access alone does not establish a working private path. The preventive control that addresses this condition is: Maintain a documented injection-regression suite and forbid unreviewed elevation of retrieved content into trusted instructions.
Question 441A financial analysis assistant includes vendor-provided web snippets that contain instructions unrelated to the user's accounting question. What is the best operational configuration or action for this situation? Version prompt templates with placeholders and apply approved secret and personal-data handling rules rather than committing live records. Keep untrusted retrieved text and user content distinct from system instructions and test prompt-injection attempts during prompt design. Configure a repository workflow that validates prompt syntax, required variables, and evaluation gates before deploying prompt changes. Evaluate provisioned throughput units for a predictable high-volume inference workload and size deployment capacity to measured token demand. ExplanationCorrect Answer B:Keep untrusted retrieved text and user content distinct from system instructions and test prompt-injection attempts during prompt design. This is appropriate because Retrieved documents can contain adversarial instructions; instruction hierarchy and explicit context boundaries reduce the chance of treating data as directives. Verify with: Evaluate representative benign retrievals and synthetic instruction-injection payloads against the same prompt version. Other Answers A:Prompt source control should contain reusable instructions, not private customer inputs or API credentials that can leak through repository history. The required action here is: Keep untrusted retrieved text and user content distinct from system instructions and test prompt-injection attempts during prompt design. C:Treating prompt assets as deployable artifacts enables repeatable validation and prevents malformed placeholders from reaching applications. The required action here is: Keep untrusted retrieved text and user content distinct from system instructions and test prompt-injection attempts during prompt design. D:Provisioned throughput can reserve model serving capacity under supported Azure offerings, unlike purely variable shared-capacity consumption. The required action here is: Keep untrusted retrieved text and user content distinct from system instructions and test prompt-injection attempts during prompt design.
Question 442A financial analysis assistant includes vendor-provided web snippets that contain instructions unrelated to the user's accounting question. Which platform mechanism explains the proposed solution? Consistent input cases and metrics reveal whether a prompt revision improves outcomes rather than relying on a few favorable examples. Retrieved documents can contain adversarial instructions; instruction hierarchy and explicit context boundaries reduce the chance of treating data as directives. Treating prompt assets as deployable artifacts enables repeatable validation and prevents malformed placeholders from reaching applications. HTTP 429 responses signal request or token throughput limits; immediate synchronized retries can make overload worse. ExplanationCorrect Answer B:Retrieved documents can contain adversarial instructions; instruction hierarchy and explicit context boundaries reduce the chance of treating data as directives. This explains the relevant intervention: Keep untrusted retrieved text and user content distinct from system instructions and test prompt-injection attempts during prompt design. Confirm by: Evaluate representative benign retrievals and synthetic instruction-injection payloads against the same prompt version. Other Answers A:Consistent input cases and metrics reveal whether a prompt revision improves outcomes rather than relying on a few favorable examples. The more specific mechanism behind the reported outcome is: Retrieved documents can contain adversarial instructions; instruction hierarchy and explicit context boundaries reduce the chance of treating data as directives. C:Treating prompt assets as deployable artifacts enables repeatable validation and prevents malformed placeholders from reaching applications. The more specific mechanism behind the reported outcome is: Retrieved documents can contain adversarial instructions; instruction hierarchy and explicit context boundaries reduce the chance of treating data as directives. D:HTTP 429 responses signal request or token throughput limits; immediate synchronized retries can make overload worse. The more specific mechanism behind the reported outcome is: Retrieved documents can contain adversarial instructions; instruction hierarchy and explicit context boundaries reduce the chance of treating data as directives.
Question 443A financial analysis assistant includes vendor-provided web snippets that contain instructions unrelated to the user's accounting question. How should this change be validated in a safe pilot? Run the same evaluation dataset through baseline and candidate prompts and record the per-case score changes. Submit a deliberately invalid prompt template in a test branch and verify CI blocks deployment with a clear diagnostic. Run a controlled throttling simulation and confirm retry limits, wait intervals, and eventual degradation behavior. Evaluate representative benign retrievals and synthetic instruction-injection payloads against the same prompt version. ExplanationCorrect Answer D:Evaluate representative benign retrievals and synthetic instruction-injection payloads against the same prompt version. This directly tests whether the intervention works: Keep untrusted retrieved text and user content distinct from system instructions and test prompt-injection attempts during prompt design. The technical reason is: Retrieved documents can contain adversarial instructions; instruction hierarchy and explicit context boundaries reduce the chance of treating data as directives. Other Answers A:Consistent input cases and metrics reveal whether a prompt revision improves outcomes rather than relying on a few favorable examples. The applicable evidence for this requirement is: Evaluate representative benign retrievals and synthetic instruction-injection payloads against the same prompt version. B:Treating prompt assets as deployable artifacts enables repeatable validation and prevents malformed placeholders from reaching applications. The applicable evidence for this requirement is: Evaluate representative benign retrievals and synthetic instruction-injection payloads against the same prompt version. C:HTTP 429 responses signal request or token throughput limits; immediate synchronized retries can make overload worse. The applicable evidence for this requirement is: Evaluate representative benign retrievals and synthetic instruction-injection payloads against the same prompt version.
Question 444A financial analysis assistant includes vendor-provided web snippets that contain instructions unrelated to the user's accounting question. What change-management control best protects this workflow? Enforce prompt repository secret scanning and fixture anonymization before review and deployment. Maintain a documented injection-regression suite and forbid unreviewed elevation of retrieved content into trusted instructions. Use protected branches and pinned prompt versions for controlled rollout and rollback. Document the selected serving model's cost, support boundaries, and change-management implications. ExplanationCorrect Answer B:Maintain a documented injection-regression suite and forbid unreviewed elevation of retrieved content into trusted instructions. This reduces the failure risk because Retrieved documents can contain adversarial instructions; instruction hierarchy and explicit context boundaries reduce the chance of treating data as directives. Immediate action: Keep untrusted retrieved text and user content distinct from system instructions and test prompt-injection attempts during prompt design. Other Answers A:Prompt source control should contain reusable instructions, not private customer inputs or API credentials that can leak through repository history. The preventive control that addresses this condition is: Maintain a documented injection-regression suite and forbid unreviewed elevation of retrieved content into trusted instructions. C:Treating prompt assets as deployable artifacts enables repeatable validation and prevents malformed placeholders from reaching applications. The preventive control that addresses this condition is: Maintain a documented injection-regression suite and forbid unreviewed elevation of retrieved content into trusted instructions. D:Serverless inference reduces infrastructure management, whereas managed compute can provide needed customization or deployment control where supported. The preventive control that addresses this condition is: Maintain a documented injection-regression suite and forbid unreviewed elevation of retrieved content into trusted instructions.
Question 445A developer pastes customer medical records into a system prompt example committed to an internal Git repository. What is the most defensible change for the stated failure? Version prompt templates with placeholders and apply approved secret and personal-data handling rules rather than committing live records. Configure a repository workflow that validates prompt syntax, required variables, and evaluation gates before deploying prompt changes. Store production system prompts and configuration templates in a versioned repository with reviewed release tags. Provision Foundry resources, project configuration, and related identity roles from reviewed Bicep templates and Azure CLI commands. ExplanationCorrect Answer A:Version prompt templates with placeholders and apply approved secret and personal-data handling rules rather than committing live records. This is appropriate because Prompt source control should contain reusable instructions, not private customer inputs or API credentials that can leak through repository history. Verify with: Inspect the published prompt artifact for placeholders, redacted fixtures, and absence of embedded credentials. Other Answers B:Treating prompt assets as deployable artifacts enables repeatable validation and prevents malformed placeholders from reaching applications. The required action here is: Version prompt templates with placeholders and apply approved secret and personal-data handling rules rather than committing live records. C:Prompt text is application behavior; version control enables traceable changes, reviews, and exact re-deployment of prior versions. The required action here is: Version prompt templates with placeholders and apply approved secret and personal-data handling rules rather than committing live records. D:Declarative infrastructure captures repeatable resource definitions and makes environmental drift reviewable through version control. The required action here is: Version prompt templates with placeholders and apply approved secret and personal-data handling rules rather than committing live records.
Question 446A developer pastes customer medical records into a system prompt example committed to an internal Git repository. Why does the recommended approach address the observed issue? Prompt source control should contain reusable instructions, not private customer inputs or API credentials that can leak through repository history. Treating prompt assets as deployable artifacts enables repeatable validation and prevents malformed placeholders from reaching applications. Retrieved documents can contain adversarial instructions; instruction hierarchy and explicit context boundaries reduce the chance of treating data as directives. Private endpoints require consistent name resolution and routing for approved resources; blocking public access alone does not establish a working private path. ExplanationCorrect Answer A:Prompt source control should contain reusable instructions, not private customer inputs or API credentials that can leak through repository history. This explains the relevant intervention: Version prompt templates with placeholders and apply approved secret and personal-data handling rules rather than committing live records. Confirm by: Inspect the published prompt artifact for placeholders, redacted fixtures, and absence of embedded credentials. Other Answers B:Treating prompt assets as deployable artifacts enables repeatable validation and prevents malformed placeholders from reaching applications. The more specific mechanism behind the reported outcome is: Prompt source control should contain reusable instructions, not private customer inputs or API credentials that can leak through repository history. C:Retrieved documents can contain adversarial instructions; instruction hierarchy and explicit context boundaries reduce the chance of treating data as directives. The more specific mechanism behind the reported outcome is: Prompt source control should contain reusable instructions, not private customer inputs or API credentials that can leak through repository history. D:Private endpoints require consistent name resolution and routing for approved resources; blocking public access alone does not establish a working private path. The more specific mechanism behind the reported outcome is: Prompt source control should contain reusable instructions, not private customer inputs or API credentials that can leak through repository history.
Question 447A developer pastes customer medical records into a system prompt example committed to an internal Git repository. What observable evidence should the engineering team collect? Inspect the published prompt artifact for placeholders, redacted fixtures, and absence of embedded credentials. Evaluate representative benign retrievals and synthetic instruction-injection payloads against the same prompt version. Retrieve a pinned prompt commit and confirm the deployed assistant uses the expected content and model settings. Run a sandbox retrieval call under the deployed identity and verify only the intended data index or storage scope is accessible. ExplanationCorrect Answer A:Inspect the published prompt artifact for placeholders, redacted fixtures, and absence of embedded credentials. This directly tests whether the intervention works: Version prompt templates with placeholders and apply approved secret and personal-data handling rules rather than committing live records. The technical reason is: Prompt source control should contain reusable instructions, not private customer inputs or API credentials that can leak through repository history. Other Answers B:Retrieved documents can contain adversarial instructions; instruction hierarchy and explicit context boundaries reduce the chance of treating data as directives. The applicable evidence for this requirement is: Inspect the published prompt artifact for placeholders, redacted fixtures, and absence of embedded credentials. C:Prompt text is application behavior; version control enables traceable changes, reviews, and exact re-deployment of prior versions. The applicable evidence for this requirement is: Inspect the published prompt artifact for placeholders, redacted fixtures, and absence of embedded credentials. D:Managed identities enable Azure service-to-service authorization using role assignments and token acquisition without distributing reusable secrets. The applicable evidence for this requirement is: Inspect the published prompt artifact for placeholders, redacted fixtures, and absence of embedded credentials.
Question 448A developer pastes customer medical records into a system prompt example committed to an internal Git repository. Which documented safeguard helps sustain the improvement? Enforce prompt repository secret scanning and fixture anonymization before review and deployment. Use protected branches and pinned prompt versions for controlled rollout and rollback. Retain a regression test dataset and review prompt variants before merging them into the production branch. Define version promotion gates and reversible rollout runbooks with a named on-call owner. ExplanationCorrect Answer A:Enforce prompt repository secret scanning and fixture anonymization before review and deployment. This reduces the failure risk because Prompt source control should contain reusable instructions, not private customer inputs or API credentials that can leak through repository history. Immediate action: Version prompt templates with placeholders and apply approved secret and personal-data handling rules rather than committing live records. Other Answers B:Treating prompt assets as deployable artifacts enables repeatable validation and prevents malformed placeholders from reaching applications. The preventive control that addresses this condition is: Enforce prompt repository secret scanning and fixture anonymization before review and deployment. C:Consistent input cases and metrics reveal whether a prompt revision improves outcomes rather than relying on a few favorable examples. The preventive control that addresses this condition is: Enforce prompt repository secret scanning and fixture anonymization before review and deployment. D:Deployment aliases and routed application settings can isolate candidate versions, allowing rollback without reconstructing the entire service. The preventive control that addresses this condition is: Enforce prompt repository secret scanning and fixture anonymization before review and deployment.
Question 449A university assistant is judged by five short demo questions even though users ask hundreds of different policy queries. Which implementation choice most directly addresses the described requirement? Build a representative held-out evaluation dataset with mapped inputs, expected outcomes, and segment annotations for the target application. Evaluate whether generated claims are supported by the retrieved reference context, separately from general fluency. Implement an application-specific evaluator that checks required schema fields and business rules in addition to general LLM quality metrics. Monitor request latency percentiles, throughput, throttling, and error rates for the production Foundry model deployment. ExplanationCorrect Answer A:Build a representative held-out evaluation dataset with mapped inputs, expected outcomes, and segment annotations for the target application. This is appropriate because Evaluation quality depends on test cases reflecting real task variety and on correct mapping between prompts, references, and outputs. Verify with: Run the evaluator on a small labeled sample and inspect that each score is attached to its intended input and expected answer. Other Answers B:Groundedness measures how well an answer is justified by supplied evidence; a fluent response may still invent unsupported facts. The required action here is: Build a representative held-out evaluation dataset with mapped inputs, expected outcomes, and segment annotations for the target application. C:Generic fluency and relevance metrics do not guarantee a JSON response satisfies a downstream API contract. The required action here is: Build a representative held-out evaluation dataset with mapped inputs, expected outcomes, and segment annotations for the target application. D:Average response latency can hide severe tail delays; throughput and throttling metrics reveal serving bottlenecks. The required action here is: Build a representative held-out evaluation dataset with mapped inputs, expected outcomes, and segment annotations for the target application.
Question 450A university assistant is judged by five short demo questions even though users ask hundreds of different policy queries. What technical principle makes this decision suitable? Evaluation quality depends on test cases reflecting real task variety and on correct mapping between prompts, references, and outputs. Generic fluency and relevance metrics do not guarantee a JSON response satisfies a downstream API contract. A model can score well on relevance while still returning unsafe content; safety testing is a separate release requirement. Token-based usage explains expenditure changes that raw request counts may conceal when prompt length varies. ExplanationCorrect Answer A:Evaluation quality depends on test cases reflecting real task variety and on correct mapping between prompts, references, and outputs. This explains the relevant intervention: Build a representative held-out evaluation dataset with mapped inputs, expected outcomes, and segment annotations for the target application. Confirm by: Run the evaluator on a small labeled sample and inspect that each score is attached to its intended input and expected answer. Other Answers B:Generic fluency and relevance metrics do not guarantee a JSON response satisfies a downstream API contract. The more specific mechanism behind the reported outcome is: Evaluation quality depends on test cases reflecting real task variety and on correct mapping between prompts, references, and outputs. C:A model can score well on relevance while still returning unsafe content; safety testing is a separate release requirement. The more specific mechanism behind the reported outcome is: Evaluation quality depends on test cases reflecting real task variety and on correct mapping between prompts, references, and outputs. D:Token-based usage explains expenditure changes that raw request counts may conceal when prompt length varies. The more specific mechanism behind the reported outcome is: Evaluation quality depends on test cases reflecting real task variety and on correct mapping between prompts, references, and outputs.