Microsoft AI-300 Practice Test 36 - Prepaway
Practice Exams:

Pass Microsoft AI-300 Exam in First Attempt Guaranteed!

Get 100% Latest Exam Questions, Accurate & Verified Answers!

30 Days Free Updates, Instant Download!

Verified By Experts
Premium File
$69.99 $76.99

AI-300 Premium ETE File


  • Premium File 213 Questions & Answers. Last Update: Nov 12, 2026
Whats Included:
  • Latest Questions
  • 100% Accurate Answers
  • Fast Exam Updates

Microsoft AI-300 Practice Test 36

Question 350

While operating claims decision support and policy analysis at a national insurance service, administrators verify this issue: A generative assistant flags harmless educational content while missing clearly prohibited harmful requests. Investigators found: Evaluator configurations use unsuitable generic thresholds and no review of the intended safety policy categories. They propose to calibrate supported safety evaluators using approved safe and unsafe benchmark prompts with human policy review. Which technical principle most directly justifies the plan?

  1. Generative AI evaluation quality is limited by representative coverage of real tasks and negative cases
  2. Safety evaluation requires policy-aligned thresholds and measured tradeoffs, not a single untested score
  3. Answer relevance and factual support measure different qualities and should not be treated as interchangeable
  4. A generative system can be technically available while producing poor-quality answers

Explanation

Correct Answer

B:

Safety evaluation requires policy-aligned thresholds and measured tradeoffs, not a single untested score. Evaluator configurations use unsuitable generic thresholds and no review of the intended safety policy categories. Human policy review with False positives and false negatives meet documented tolerances on a held-out safety set.

Incorrect Answers

A:

Generative AI evaluation quality is limited by representative coverage of real tasks and negative cases. Safety evaluation requires policy-aligned thresholds and measured tradeoffs, not a single untested score.

C:

Answer relevance and factual support measure different qualities and should not be treated as interchangeable. Evaluator configurations use unsuitable generic thresholds and no review of the intended safety policy categories. Safety evaluation requires policy-aligned thresholds and measured tradeoffs, not a single untested score.

D:

A generative system can be technically available while producing poor-quality answers. Evaluator configurations use unsuitable generic thresholds and no review of the intended safety policy categories. Safety evaluation requires policy-aligned thresholds and measured tradeoffs, not a single untested score.

Question 351

A controlled Azure AI systems audit at an international education provider focuses on tutoring assistants and course administration. It confirms that An automated provisioning run creates a production Machine Learning workspace outside the approved data residency region. The proposed fix is to set the explicit approved location in the reviewed Bicep deployment and rebuild the workspace with validated dependent resources. Preserve sensitive data, approved model artifacts and least-privilege controls during the test. Which observation most decisively establishes the stated root cause?

  1. Storage diagnostic logs identify denied data-plane access from the job or workspace managed identity
  2. Private networking assessment shows public access enabled although the project requires a restricted workspace
  3. Deployment history and workspace location properties show the resource group location was assumed to control the workspace's actual region
  4. DNS queries from the managed runtime return public service records while required private DNS links are missing

Explanation

Correct Answer

C:

A controlled Azure AI systems audit at an international education provider focuses on tutoring assistants and course administration. Deployment history and workspace location properties show the resource group location was assumed to control the workspace's actual region. An automated provisioning run creates a production Machine Learning workspace outside the approved data residency region. Set the explicit approved location in the reviewed Bicep deployment and rebuild the workspace with validated dependent resources.

Incorrect Answers

A:

Storage diagnostic logs identify denied data-plane access from the job or workspace managed identity. Deployment history and workspace location properties show the resource group location was assumed to control the workspace's actual region. Set the explicit approved location in the reviewed Bicep deployment and rebuild the workspace with validated dependent resources.

B:

Private networking assessment shows public access enabled although the project requires a restricted workspace. Set the explicit approved location in the reviewed Bicep deployment and rebuild the workspace with validated dependent resources.

D:

DNS queries from the managed runtime return public service records while required private DNS links are missing. Deployment history and workspace location properties show the resource group location was assumed to control the workspace's actual region. Set the explicit approved location in the reviewed Bicep deployment and rebuild the workspace with validated dependent resources.

Question 352

During an approved Azure AI operations review at a transport authority, the engineering team supporting timetable reliability and passenger assistance discovers that An agent client retries throttled completions immediately and compounds service failures during a burst. The approved change is to implement bounded exponential backoff with jitter and a workload-appropriate concurrency budget. Validate the change in a governed canary before production promotion. Which post-change measurable result best validates success?

  1. A representative peak load meets token throughput and latency objectives without excessive throttling
  2. Required languages meet minimum correctness and safety acceptance without reducing compliant English performance
  3. Approved test prompts preserve expected behavior without personal records in shared source control
  4. A controlled throttle test recovers without runaway request amplification and meets retry budgets

Explanation

Correct Answer

D:

A controlled throttle test recovers without runaway request amplification and meets retry budgets. Telemetry shows correlated 429 errors and identical parallel retry spikes with no jitter or concurrency ceiling. Implement bounded exponential backoff with jitter and a workload-appropriate concurrency budget.

Incorrect Answers

A:

A representative peak load meets token throughput and latency objectives without excessive throttling. Telemetry shows correlated 429 errors and identical parallel retry spikes with no jitter or concurrency ceiling. A controlled throttle test recovers without runaway request amplification and meets retry budgets.

B:

Required languages meet minimum correctness and safety acceptance without reducing compliant English performance. Telemetry shows correlated 429 errors and identical parallel retry spikes with no jitter or concurrency ceiling. A controlled throttle test recovers without runaway request amplification and meets retry budgets.

C:

Approved test prompts preserve expected behavior without personal records in shared source control. A controlled throttle test recovers without runaway request amplification and meets retry budgets.

Question 353

During an approved Azure AI operations review at a logistics network, the engineering team supporting fleet ETA forecasting and supply chain support discovers that A classification model passes aggregate accuracy review while error rates are sharply worse for a minority cohort. Diagnostic evidence: Subgroup confusion matrices reveal unequal false-negative rates hidden by a high overall accuracy score. Validate the change in a governed canary before production promotion. Which action together with its verification criterion is most appropriate?

  1. Add appropriate approved responsible AI subgroup metrics and mitigation review before model release — verification: Evaluation reports both aggregate and relevant subgroup performance against documented thresholds
  2. Register the evaluated artifact from the approved run and attach the correct run and data version lineage — verification: A reviewer can traverse the registered model to the validated run parameters, metrics and immutable dataset
  3. Package the approved feature retrieval definition and input contract with the model version before registration — verification: A fresh deployment reconstructs feature lookup and passes train-versus-serve feature parity tests
  4. Select a Microsoft-supported GPU runtime image and validate framework-driver compatibility on the target compute — verification: A reproducible smoke test detects the GPU and completes a short forward-and-backward pass

Explanation

Correct Answer

A:

Add appropriate approved responsible AI subgroup metrics and mitigation review before model release. Evaluation reports both aggregate and relevant subgroup performance against documented thresholds. Subgroup confusion matrices reveal unequal false-negative rates hidden by a high overall accuracy score. Responsible ML release gates must inspect material cohort harms rather than rely exclusively on average accuracy.

Incorrect Answers

B:

Register the evaluated artifact from the approved run and attach the correct run and data version lineage. A reviewer can traverse the registered model to the validated run parameters, metrics and immutable dataset. This pair concerns a different control or outcome, leaving the recorded fault unresolved. Add appropriate approved responsible AI subgroup metrics and mitigation review before model release; Evaluation reports both aggregate and relevant subgroup performance against documented thresholds.

C:

Package the approved feature retrieval definition and input contract with the model version before registration. A fresh deployment reconstructs feature lookup and passes train-versus-serve feature parity tests. Add appropriate approved responsible AI subgroup metrics and mitigation review before model release; Evaluation reports both aggregate and relevant subgroup performance against documented thresholds.

D:

Select a Microsoft-supported GPU runtime image and validate framework-driver compatibility on the target compute. A reproducible smoke test detects the GPU and completes a short forward-and-backward pass. Add appropriate approved responsible AI subgroup metrics and mitigation review before model release; Evaluation reports both aggregate and relevant subgroup performance against documented thresholds.

Question 354

Users of tutoring assistants and course administration at an international education provider report a machine learning or agent operations problem: An assistant model chosen for an international call center passes English validation but degrades for supported local languages. The evidence is: Version comparison evaluation shows significantly worse relevant-language responses despite improved English score. Avoid an unrelated subscription-wide identity, networking or model change. Which corrective action most directly addresses the failure?

  1. Retain the last validated version and gate traffic shifts on schema and agent integration tests
  2. Select or tune the candidate using representative approved multilingual evaluation sets and documented constraints
  3. Select a regionally supported model deployment whose advertised API features meet the agent contract and retest tools
  4. Implement bounded exponential backoff with jitter and a workload-appropriate concurrency budget

Explanation

Correct Answer

B:

Select or tune the candidate using representative approved multilingual evaluation sets and documented constraints. Version comparison evaluation shows significantly worse relevant-language responses despite improved English score. Required languages meet minimum correctness and safety acceptance without reducing compliant English performance.

Incorrect Answers

A:

Retain the last validated version and gate traffic shifts on schema and agent integration tests. Version comparison evaluation shows significantly worse relevant-language responses despite improved English score. Select or tune the candidate using representative approved multilingual evaluation sets and documented constraints.

C:

Select a regionally supported model deployment whose advertised API features meet the agent contract and retest tools. Version comparison evaluation shows significantly worse relevant-language responses despite improved English score. Select or tune the candidate using representative approved multilingual evaluation sets and documented constraints.

D:

Implement bounded exponential backoff with jitter and a workload-appropriate concurrency budget. Version comparison evaluation shows significantly worse relevant-language responses despite improved English score. Select or tune the candidate using representative approved multilingual evaluation sets and documented constraints.

Question 355

The MLOps and GenAIOps team at a logistics network operates fleet ETA forecasting and supply chain support. An incident investigation establishes that An agent appears to spend all response time in model inference because retrieval tool spans are missing from distributed traces. The approved change is to enable supported distributed tracing across retrieval and generation with stable correlation identifiers. Keep changes within the authorized Azure ML or Foundry project and its data access boundaries. Which post-change measurable result best validates success?

  1. Traces still explain a controlled failure but sampled output contains no unauthorized sensitive attributes
  2. A controlled tool failure triggers the correct alert and is correlated to an affected user request
  3. An end-to-end test trace shows tool latency, model latency and response time without storing sensitive document text
  4. Every required intent category has a test and the release gate rejects a regression in account recovery

Explanation

Correct Answer

C:

An end-to-end test trace shows tool latency, model latency and response time without storing sensitive document text. Useful agent observability requires correlated multi-component traces rather than model-only telemetry.

Incorrect Answers

A:

Traces still explain a controlled failure but sampled output contains no unauthorized sensitive attributes. An end-to-end test trace shows tool latency, model latency and response time without storing sensitive document text. Useful agent observability requires correlated multi-component traces rather than model-only telemetry.

B:

A controlled tool failure triggers the correct alert and is correlated to an affected user request. An end-to-end test trace shows tool latency, model latency and response time without storing sensitive document text.

D:

Every required intent category has a test and the release gate rejects a regression in account recovery. An end-to-end test trace shows tool latency, model latency and response time without storing sensitive document text.

Question 356

At a digital commerce operator, an AI model release board examines a failure affecting fraud monitoring and shopping assistance: An agent release fails when a prompt template expects a customer_context field that the application no longer supplies. The approved change is to version the template variable contract and validate input fields in the same application release pipeline. Avoid an unrelated subscription-wide identity, networking or model change. Which post-change measurable result best validates success?

  1. A clean release reproduces the approved prompt and an unreviewed portal edit is detected
  2. A restored deployment passes the prior acceptance set with matched prompt and model versions
  3. A failed canary reverts to the retained endpoint and previously supported tool actions remain valid
  4. Approved requests render the intended prompt and missing fields fail validation before model invocation

Explanation

Correct Answer

D:

Approved requests render the intended prompt and missing fields fail validation before model invocation. An agent release fails when a prompt template expects a customer_context field that the application no longer supplies. Version the template variable contract and validate input fields in the same application release pipeline.

Incorrect Answers

A:

A clean release reproduces the approved prompt and an unreviewed portal edit is detected. Approved requests render the intended prompt and missing fields fail validation before model invocation.

B:

A restored deployment passes the prior acceptance set with matched prompt and model versions. An agent release fails when a prompt template expects a customer_context field that the application no longer supplies. Approved requests render the intended prompt and missing fields fail validation before model invocation.

C:

A failed canary reverts to the retained endpoint and previously supported tool actions remain valid. An agent release fails when a prompt template expects a customer_context field that the application no longer supplies. Approved requests render the intended prompt and missing fields fail validation before model invocation.

Question 357

A controlled Azure AI systems audit at a field services franchise focuses on technician dispatch and service-history retrieval. It confirms that A compliance assistant fails to find a document with an exact regulation code because semantic similarity favors broader summaries. The evidence is: Retrieval traces show keyword-specific identifier queries ranking unrelated conceptually similar content above the exact match. Preserve sensitive data, approved model artifacts and least-privilege controls during the test. Which corrective action most directly addresses the failure?

  1. Use appropriately configured hybrid lexical and vector retrieval with measured weighting and supported reranking
  2. Align indexing and query embeddings to the same approved model version and reindex under controlled rollout
  3. Evaluate retrieval precision, groundedness and answer support using known query-document reference pairs
  4. Filter synthetic records against approved source policy and test factual outputs on an independent set

Explanation

Correct Answer

A:

Use appropriately configured hybrid lexical and vector retrieval with measured weighting and supported reranking. Retrieval traces show keyword-specific identifier queries ranking unrelated conceptually similar content above the exact match. Hybrid search can recover lexical exact-match signals that pure embeddings sometimes dilute.

Incorrect Answers

B:

Align indexing and query embeddings to the same approved model version and reindex under controlled rollout. Use appropriately configured hybrid lexical and vector retrieval with measured weighting and supported reranking.

C:

Evaluate retrieval precision, groundedness and answer support using known query-document reference pairs. Retrieval traces show keyword-specific identifier queries ranking unrelated conceptually similar content above the exact match. Use appropriately configured hybrid lexical and vector retrieval with measured weighting and supported reranking.

D:

Filter synthetic records against approved source policy and test factual outputs on an independent set. Retrieval traces show keyword-specific identifier queries ranking unrelated conceptually similar content above the exact match. Use appropriately configured hybrid lexical and vector retrieval with measured weighting and supported reranking.

Question 358

At a biotech research organization, an AI model release board examines a failure affecting laboratory observations and clinical data summaries: A research notebook gives good accuracy but a pipeline job cannot find the same feature transformations or input file. The evidence is: Notebook execution history references a local path and interactive in-memory state outside the committed pipeline. Avoid an unrelated subscription-wide identity, networking or model change. Which corrective action most directly addresses the failure?

  1. Set justified sampling, early termination and compute limits that permit the approved search budget
  2. Package feature transformations as versioned components and register required data assets with reproducible inputs
  3. Version the partition definition and set deterministic approved splitting for reproducible model comparisons
  4. Calibrate per-cohort thresholds and require appropriate quality signals or human approval before retraining

Explanation

Correct Answer

B:

Package feature transformations as versioned components and register required data assets with reproducible inputs. Notebook execution history references a local path and interactive in-memory state outside the committed pipeline. Reproducible inputs with A fresh noninteractive run on clean compute reproduces the evaluated transformations and performance range.

Incorrect Answers

A:

Set justified sampling, early termination and compute limits that permit the approved search budget. Notebook execution history references a local path and interactive in-memory state outside the committed pipeline. Reproducible inputs with A fresh noninteractive run on clean compute reproduces the evaluated transformations and performance range.

C:

Version the partition definition and set deterministic approved splitting for reproducible model comparisons. Reproducible inputs with A fresh noninteractive run on clean compute reproduces the evaluated transformations and performance range.

D:

Calibrate per-cohort thresholds and require appropriate quality signals or human approval before retraining. Notebook execution history references a local path and interactive in-memory state outside the committed pipeline. Reproducible inputs with A fresh noninteractive run on clean compute reproduces the evaluated transformations and performance range.

Question 359

While operating patient-support triage and clinical scheduling at a regional healthcare provider, administrators verify this issue: A prompt engineer includes real customer examples in a versioned prompt demonstration that is later shared externally. The proposed fix is to remove and remediate exposed examples, substitute approved anonymized fixtures and control repository visibility. Do not grant broader model, dataset, or infrastructure permissions than the approved operation requires. Which observation most decisively establishes the stated root cause?

  1. Prompt artifact checksum differs from Git and there is no approved change associated with the live revision
  2. Release manifests show the restored prompt digest paired with an untested model deployment version
  3. Repository scan finds protected names and records in example messages despite no API keys present
  4. GitHub environment variables and deployment logs show a stale project endpoint identifier in the protected branch

Explanation

Correct Answer

C:

Repository scan finds protected names and records in example messages despite no API keys present. A prompt engineer includes real customer examples in a versioned prompt demonstration that is later shared externally. Remove and remediate exposed examples, substitute approved anonymized fixtures and control repository visibility.

Incorrect Answers

A:

Prompt artifact checksum differs from Git and there is no approved change associated with the live revision. Repository scan finds protected names and records in example messages despite no API keys present. Remove and remediate exposed examples, substitute approved anonymized fixtures and control repository visibility.

B:

Release manifests show the restored prompt digest paired with an untested model deployment version. Repository scan finds protected names and records in example messages despite no API keys present. Remove and remediate exposed examples, substitute approved anonymized fixtures and control repository visibility.

D:

GitHub environment variables and deployment logs show a stale project endpoint identifier in the protected branch. Remove and remediate exposed examples, substitute approved anonymized fixtures and control repository visibility.

Question 360

During an approved Azure AI operations review at a technology consulting firm, the engineering team supporting customer advisory agents and model deployment services discovers that A scheduled training pipeline remains queued indefinitely despite a successful submitted job record. The approved change is to choose a supported available SKU or request the required capacity quota through approved provisioning processes. Validate the change in a governed canary before production promotion. Which post-change measurable result best validates success?

  1. A fresh job retrieves the intended versioned dataset without exposing storage account keys
  2. Nightly training starts on provisioned job compute even when the developer instance is stopped
  3. The pipeline compiles and outputs the expected named asset without editing already approved old components
  4. A test job allocates the required compute and completes within the approved training start SLO

Explanation

Correct Answer

D:

A test job allocates the required compute and completes within the approved training start SLO. A scheduled training pipeline remains queued indefinitely despite a successful submitted job record. Choose a supported available SKU or request the required capacity quota through approved provisioning processes.

Incorrect Answers

A:

A fresh job retrieves the intended versioned dataset without exposing storage account keys. A scheduled training pipeline remains queued indefinitely despite a successful submitted job record. A test job allocates the required compute and completes within the approved training start SLO.

B:

Nightly training starts on provisioned job compute even when the developer instance is stopped. A scheduled training pipeline remains queued indefinitely despite a successful submitted job record. A test job allocates the required compute and completes within the approved training start SLO.

C:

The pipeline compiles and outputs the expected named asset without editing already approved old components. A scheduled training pipeline remains queued indefinitely despite a successful submitted job record. A test job allocates the required compute and completes within the approved training start SLO.

Operationalizing Machine Learning and Generative AI Solutions (AI-300) Practice Tests