Practice Exams:

Microsoft AI-103 Dumps

EXAM: Developing AI Apps and Agents on Azure
Verified By Experts
Premium Bundle
$49.99

AI-103 Premium ETE File


  • Premium File 187 Questions & Answers
Whats Included:
  • Latest Questions
  • 100% Accurate Answers
  • Fast Exam Updates

Microsoft AI-103 Practice Test 9

Question 80

During a release acceptance review, a responsible-AI reviewer must accept a change for foundry project isolation. The issue to be addressed is: Two business units must administer AI resources without seeing one another's experiments. Which concrete verification action demonstrates that the relevant behavior works?

  1. Invoke the endpoint through the deployed workload identity and inspect denied extra privileges
  2. Grant both teams unrestricted project-owner access to every Foundry resource
  3. Create appropriately scoped projects and RBAC assignments for each business unit
  4. Sign in as each team and verify permitted projects and denied unrelated projects

Explanation

Correct Answer

D:

To verify this change, the insurance platform should sign in as each team and verify permitted projects and denied unrelated projects. Operators still need the implementation to create appropriately scoped projects and RBAC assignments for each business unit and the preventive requirement to define project ownership and least-privilege access reviews.

Incorrect Answers

A:

For managed identity model access, it can be sensible to invoke the endpoint through the deployed workload identity and inspect denied extra privileges if an app must call a Foundry inference endpoint without embedding service secrets. Without the actual change to sign in as each team and verify permitted projects and denied unrelated projects, the answer lacks the correct operational scope even if the alternate action is technically valid elsewhere.

B:

In the insurance platform acceptance review, trying to grant both teams unrestricted project-owner access to every Foundry resource does not test the exact foundry project isolation condition where two business units must administer AI resources without seeing one another's experiments; the relevant test is to sign in as each team and verify permitted projects and denied unrelated projects. Acceptance should instead demonstrate that the service can sign in as each team and verify permitted projects and denied unrelated projects, using the described failure as the test criterion.

C:

The action to create appropriately scoped projects and RBAC assignments for each business unit can contribute to foundry project isolation as a implementation step; here, two business units must administer AI resources without seeing one another's experiments. In the insurance platform acceptance review, trying to create appropriately scoped projects and RBAC assignments for each business unit does not test the exact foundry project isolation condition where two business units must administer AI resources without seeing one another's experiments; the relevant test is to sign in as each team and verify permitted projects and denied unrelated projects.

Question 81

At a research university, the solution architect is preparing a quality regression test. The concrete requirement is: An app must call a Foundry inference endpoint without embedding service secrets. Which design action BEST addresses the stated need?

  1. Store long-lived service API keys in a public configuration repository
  2. Use supported Microsoft Entra managed-identity authorization with least-privilege roles
  3. Invoke the endpoint through the deployed workload identity and inspect denied extra privileges
  4. Create appropriately scoped projects and RBAC assignments for each business unit

Explanation

Correct Answer

B:

The proposed implementation is to use supported Microsoft Entra managed-identity authorization with least-privilege roles; this matches the operational requirement that an app must call a Foundry inference endpoint without embedding service secrets. The team can invoke the endpoint through the deployed workload identity and inspect denied extra privileges. The team should also forbid secrets in source control and audit identity assignments.

Incorrect Answers

A:

Choosing to store long-lived service API keys in a public configuration repository cannot directly replace the specified action to use supported Microsoft Entra managed-identity authorization with least-privilege roles when an app must call a Foundry inference endpoint without embedding service secrets. The actual implementation must use supported Microsoft Entra managed-identity authorization with least-privilege roles, because the stated service condition is that an app must call a Foundry inference endpoint without embedding service secrets.

C:

The action to invoke the endpoint through the deployed workload identity and inspect denied extra privileges can contribute to managed identity model access as a targeted acceptance check step; here, an app must call a Foundry inference endpoint without embedding service secrets. Choosing to invoke the endpoint through the deployed workload identity and inspect denied extra privileges cannot directly replace the specified action to use supported Microsoft Entra managed-identity authorization with least-privilege roles when an app must call a Foundry inference endpoint without embedding service secrets.

D:

Choosing to create appropriately scoped projects and RBAC assignments for each business unit would direct effort toward foundry project isolation, whose characteristic issue is that two business units must administer AI resources without seeing one another's experiments. The relevant corrective or preventive action here remains to use supported Microsoft Entra managed-identity authorization with least-privilege roles, with its own validation criteria.

Question 82

After a production incident at a education technology company, the team observes this failure: An app must call a Foundry inference endpoint without embedding service secrets. Which flawed assumption or implementation choice would MOST DIRECTLY explain why the intended outcome was not achieved?

  1. Invoke the endpoint through the deployed workload identity and inspect denied extra privileges
  2. Store long-lived service API keys in a public configuration repository
  3. Choose any consumption deployment without evaluating capacity predictability
  4. Use supported Microsoft Entra managed-identity authorization with least-privilege roles

Explanation

Correct Answer

B:

This design mistake is to store long-lived service API keys in a public configuration repository; it conflicts with the actual managed identity model access behavior needed when an app must call a Foundry inference endpoint without embedding service secrets. The education technology company should instead use supported Microsoft Entra managed-identity authorization with least-privilege roles, followed by a check to invoke the endpoint through the deployed workload identity and inspect denied extra privileges.

Incorrect Answers

A:

The alternative to invoke the endpoint through the deployed workload identity and inspect denied extra privileges is a targeted acceptance check step rather than the faulty assumption being diagnosed. This education technology company failure-analysis alternative of invoke the endpoint through the deployed workload identity and inspect denied extra privileges does not diagnose the particular managed identity model access defect where an app must call a Foundry inference endpoint without embedding service secrets; the keyed mistake is store long-lived service API keys in a public configuration repository.

C:

The option to choose any consumption deployment without evaluating capacity predictability would normally serve a different deployment type selection requirement, where a team needs predictable high-volume inference throughput for a stable workload. The alternative may be useful in its own context, but it does not directly supply store long-lived service API keys in a public configuration repository for the stated decision.

D:

The alternative to use supported Microsoft Entra managed-identity authorization with least-privilege roles is a implementation step rather than the faulty assumption being diagnosed. This education technology company failure-analysis alternative of use supported Microsoft Entra managed-identity authorization with least-privilege roles does not diagnose the particular managed identity model access defect where an app must call a Foundry inference endpoint without embedding service secrets; the keyed mistake is store long-lived service API keys in a public configuration repository.

Question 83

During a data protection inspection at a pharmaceutical research group, an AI team must address managed identity model access. The requirement is: An app must call a Foundry inference endpoint without embedding service secrets. Which THREE actions together implement the needed solution, its preventive guardrail, and a targeted acceptance test? (Select THREE answers.)

  1. Store long-lived service API keys in a public configuration repository
  2. Forbid secrets in source control and audit identity assignments
  3. Invoke the endpoint through the deployed workload identity and inspect denied extra privileges
  4. Replace exposed keys with workload identity permissions and rotate compromised secrets
  5. Require environment segmentation and per-stage access tests
  6. Use supported Microsoft Entra managed-identity authorization with least-privilege roles

Explanation

Correct Answers

B:

The sustained prevention measure is to forbid secrets in source control and audit identity assignments. The implementation can still use the separate action to use supported Microsoft Entra managed-identity authorization with least-privilege roles, but a guardrail constrains future changes. The pharmaceutical research group can measure whether the rule works by arranging to invoke the endpoint through the deployed workload identity and inspect denied extra privileges.

C:

The targeted acceptance test is to invoke the endpoint through the deployed workload identity and inspect denied extra privileges. The design should first use supported Microsoft Entra managed-identity authorization with least-privilege roles. A robust pharmaceutical research group deployment also needs to forbid secrets in source control and audit identity assignments; passing this check provides relevant evidence but does not certify every possible input or security risk.

F:

The proposed implementation is to use supported Microsoft Entra managed-identity authorization with least-privilege roles; this matches the operational requirement that an app must call a Foundry inference endpoint without embedding service secrets. The team can invoke the endpoint through the deployed workload identity and inspect denied extra privileges. The team should also forbid secrets in source control and audit identity assignments.

Incorrect Answers

A:

During pharmaceutical research group remediation, the proposed step to store long-lived service API keys in a public configuration repository is not the direct restoration action for managed identity model access when an app must call a Foundry inference endpoint without embedding service secrets; the keyed repair is forbid secrets in source control and audit identity assignments. After remediation the team must invoke the endpoint through the deployed workload identity and inspect denied extra privileges; the action in this option does not itself establish that target recovery outcome.

D:

The action to replace exposed keys with workload identity permissions and rotate compromised secrets can contribute to managed identity model access as a corrective remediation step; here, an app must call a Foundry inference endpoint without embedding service secrets. During pharmaceutical research group remediation, the proposed step to replace exposed keys with workload identity permissions and rotate compromised secrets is not the direct restoration action for managed identity model access when an app must call a Foundry inference endpoint without embedding service secrets; the keyed repair is forbid secrets in source control and audit identity assignments.

E:

For multi-environment data boundaries, it can be sensible to require environment segmentation and per-stage access tests if a staging agent unexpectedly accesses production customer records. Without the actual change to use supported Microsoft Entra managed-identity authorization with least-privilege roles, the answer lacks the correct operational scope even if the alternate action is technically valid elsewhere.

Question 84

A quality assurance lead at a travel support provider wants the following issue to be prevented in later deployments: An app must call a Foundry inference endpoint without embedding service secrets. Which enduring control or engineering guardrail BEST prevents recurrence?

  1. Forbid secrets in source control and audit identity assignments
  2. Store long-lived service API keys in a public configuration repository
  3. Replace exposed keys with workload identity permissions and rotate compromised secrets
  4. Validate resource IDs and authorized project scopes in deployment pipelines

Explanation

Correct Answer

A:

Standardizing the boundary to forbid secrets in source control and audit identity assignments is preventive because the same managed identity model access failure could otherwise return: an app must call a Foundry inference endpoint without embedding service secrets. For this travel support provider, the service should also use supported Microsoft Entra managed-identity authorization with least-privilege roles, and tests should invoke the endpoint through the deployed workload identity and inspect denied extra privileges so that the ongoing requirement is measurable.

Incorrect Answers

B:

For prevention at the travel support provider, the measure to store long-lived service API keys in a public configuration repository is weaker or aimed at another issue than the managed identity model access guardrail to forbid secrets in source control and audit identity assignments needed when an app must call a Foundry inference endpoint without embedding service secrets. The lasting guardrail for this specific threat is to forbid secrets in source control and audit identity assignments, not simply to complete an adjacent service task.

C:

The action to replace exposed keys with workload identity permissions and rotate compromised secrets can contribute to managed identity model access as a corrective remediation step; here, an app must call a Foundry inference endpoint without embedding service secrets. For prevention at the travel support provider, the measure to replace exposed keys with workload identity permissions and rotate compromised secrets is weaker or aimed at another issue than the managed identity model access guardrail to forbid secrets in source control and audit identity assignments needed when an app must call a Foundry inference endpoint without embedding service secrets.

D:

The action to validate resource IDs and authorized project scopes in deployment pipelines addresses foundry endpoint routing rather than the decisive managed identity model access condition. The travel support provider should not infer that fixing one adjacent AI service validates its unrelated workflow; the requested action is to forbid secrets in source control and audit identity assignments.

Question 85

During a production incident, a AI platform engineer must accept a change for managed identity model access. The issue to be addressed is: An app must call a Foundry inference endpoint without embedding service secrets. Which concrete verification action demonstrates that the relevant behavior works?

  1. Replace exposed keys with workload identity permissions and rotate compromised secrets
  2. Use supported Microsoft Entra managed-identity authorization with least-privilege roles
  3. Call inference from the application identity and verify target project logging
  4. Store long-lived service API keys in a public configuration repository
  5. Forbid secrets in source control and audit identity assignments
  6. Invoke the endpoint through the deployed workload identity and inspect denied extra privileges

Explanation

Correct Answer

F:

The targeted acceptance test is to invoke the endpoint through the deployed workload identity and inspect denied extra privileges. The design should first use supported Microsoft Entra managed-identity authorization with least-privilege roles. A robust scientific publisher deployment also needs to forbid secrets in source control and audit identity assignments; passing this check provides relevant evidence but does not certify every possible input or security risk.

Incorrect Answers

A:

The action to replace exposed keys with workload identity permissions and rotate compromised secrets can contribute to managed identity model access as a corrective remediation step; here, an app must call a Foundry inference endpoint without embedding service secrets. In the scientific publisher acceptance review, trying to replace exposed keys with workload identity permissions and rotate compromised secrets does not test the exact managed identity model access condition where an app must call a Foundry inference endpoint without embedding service secrets; the relevant test is to invoke the endpoint through the deployed workload identity and inspect denied extra privileges.

B:

The action to use supported Microsoft Entra managed-identity authorization with least-privilege roles can contribute to managed identity model access as a implementation step; here, an app must call a Foundry inference endpoint without embedding service secrets. In the scientific publisher acceptance review, trying to use supported Microsoft Entra managed-identity authorization with least-privilege roles does not test the exact managed identity model access condition where an app must call a Foundry inference endpoint without embedding service secrets; the relevant test is to invoke the endpoint through the deployed workload identity and inspect denied extra privileges.

C:

The option to call inference from the application identity and verify target project logging would normally serve a different foundry endpoint routing requirement, where an application receives authorization errors after being moved to a new project. The alternative may be useful in its own context, but it does not directly supply invoke the endpoint through the deployed workload identity and inspect denied extra privileges for the stated decision.

D:

In the scientific publisher acceptance review, trying to store long-lived service API keys in a public configuration repository does not test the exact managed identity model access condition where an app must call a Foundry inference endpoint without embedding service secrets; the relevant test is to invoke the endpoint through the deployed workload identity and inspect denied extra privileges. Acceptance should instead demonstrate that the service can invoke the endpoint through the deployed workload identity and inspect denied extra privileges, using the described failure as the test criterion.

E:

The action to forbid secrets in source control and audit identity assignments can contribute to managed identity model access as a ongoing preventive rule step; here, an app must call a Foundry inference endpoint without embedding service secrets. In the scientific publisher acceptance review, trying to forbid secrets in source control and audit identity assignments does not test the exact managed identity model access condition where an app must call a Foundry inference endpoint without embedding service secrets; the relevant test is to invoke the endpoint through the deployed workload identity and inspect denied extra privileges.

Question 86

At a insurance platform, the compliance specialist is preparing a architecture migration. The concrete requirement is: An agent deployment fails because the selected Azure region lacks model quota. Which design action BEST addresses the stated need?

  1. Implement rate-aware queueing bounded concurrency and appropriate backoff
  2. Assume the same model quota and SKU exist in every Azure region
  3. Select a model deployment region and capacity allocation that support expected demand
  4. Run sustained representative requests and verify no capacity or throttling failure

Explanation

Correct Answer

C:

Choosing to select a model deployment region and capacity allocation that support expected demand directly targets the described regional capacity planning gap: an agent deployment fails because the selected Azure region lacks model quota. As a lasting boundary, they also need to track regional quotas and expected peak request load, because an implementation by itself does not prove safe operation.

Incorrect Answers

A:

Choosing to implement rate-aware queueing bounded concurrency and appropriate backoff would direct effort toward request endpoint throttling, whose characteristic issue is that an application overwhelms an AI endpoint with simultaneous bursts. The relevant corrective or preventive action here remains to select a model deployment region and capacity allocation that support expected demand, with its own validation criteria.

B:

Choosing to assume the same model quota and SKU exist in every Azure region cannot directly replace the specified action to select a model deployment region and capacity allocation that support expected demand when an agent deployment fails because the selected Azure region lacks model quota. The actual implementation must select a model deployment region and capacity allocation that support expected demand, because the stated service condition is that an agent deployment fails because the selected Azure region lacks model quota.

D:

The action to run sustained representative requests and verify no capacity or throttling failure can contribute to regional capacity planning as a targeted acceptance check step; here, an agent deployment fails because the selected Azure region lacks model quota. Choosing to run sustained representative requests and verify no capacity or throttling failure cannot directly replace the specified action to select a model deployment region and capacity allocation that support expected demand when an agent deployment fails because the selected Azure region lacks model quota.

Question 87

After a operational readiness review at a media platform, the team observes this failure: An agent deployment fails because the selected Azure region lacks model quota. Which flawed assumption or implementation choice would MOST DIRECTLY explain why the intended outcome was not achieved?

  1. Select a model deployment region and capacity allocation that support expected demand
  2. Run sustained representative requests and verify no capacity or throttling failure
  3. Assume the same model quota and SKU exist in every Azure region
  4. Treat all model updates as identical in outputs and tool behavior

Explanation

Correct Answer

C:

The faulty assumption is assume the same model quota and SKU exist in every Azure region. A functioning alternative would be to select a model deployment region and capacity allocation that support expected demand, and the relevant validation would be to run sustained representative requests and verify no capacity or throttling failure.

Incorrect Answers

A:

The alternative to select a model deployment region and capacity allocation that support expected demand is a implementation step rather than the faulty assumption being diagnosed. This media platform failure-analysis alternative of select a model deployment region and capacity allocation that support expected demand does not diagnose the particular regional capacity planning defect where an agent deployment fails because the selected Azure region lacks model quota; the keyed mistake is assume the same model quota and SKU exist in every Azure region.

B:

The alternative to run sustained representative requests and verify no capacity or throttling failure is a targeted acceptance check step rather than the faulty assumption being diagnosed. This media platform failure-analysis alternative of run sustained representative requests and verify no capacity or throttling failure does not diagnose the particular regional capacity planning defect where an agent deployment fails because the selected Azure region lacks model quota; the keyed mistake is assume the same model quota and SKU exist in every Azure region.

D:

The option to treat all model updates as identical in outputs and tool behavior would normally serve a different model version pinning requirement, where an AI service changes behavior after a model version advances automatically. The alternative may be useful in its own context, but it does not directly supply assume the same model quota and SKU exist in every Azure region for the stated decision.

Question 88

The customer support center is responding to a regional deployment. A data engineer discovers this issue: An agent deployment fails because the selected Azure region lacks model quota. What corrective step should the team take to restore the intended functionality most directly?

  1. Assume the same model quota and SKU exist in every Azure region
  2. Correct the endpoint and assign the intended managed identity access
  3. Track regional quotas and expected peak request load
  4. Run sustained representative requests and verify no capacity or throttling failure
  5. Request suitable capacity or use an approved available deployment region

Explanation

Correct Answer

E:

The appropriate remediation is to request suitable capacity or use an approved available deployment region, because the failure concerns regional capacity planning: an agent deployment fails because the selected Azure region lacks model quota. A meaningful verification should run sustained representative requests and verify no capacity or throttling failure, and the durable operating rule should track regional quotas and expected peak request load after the immediate fix.

Incorrect Answers

A:

During customer support center remediation, the proposed step to assume the same model quota and SKU exist in every Azure region is not the direct restoration action for regional capacity planning when an agent deployment fails because the selected Azure region lacks model quota; the keyed repair is request suitable capacity or use an approved available deployment region. After remediation the team must run sustained representative requests and verify no capacity or throttling failure; the action in this option does not itself establish that target recovery outcome.

B:

The action to correct the endpoint and assign the intended managed identity access addresses foundry endpoint routing rather than the decisive regional capacity planning condition. The customer support center should not infer that fixing one adjacent AI service validates its unrelated workflow; the requested action is to request suitable capacity or use an approved available deployment region.

C:

The action to track regional quotas and expected peak request load can contribute to regional capacity planning as a ongoing preventive rule step; here, an agent deployment fails because the selected Azure region lacks model quota. During customer support center remediation, the proposed step to track regional quotas and expected peak request load is not the direct restoration action for regional capacity planning when an agent deployment fails because the selected Azure region lacks model quota; the keyed repair is request suitable capacity or use an approved available deployment region.

D:

The action to run sustained representative requests and verify no capacity or throttling failure can contribute to regional capacity planning as a targeted acceptance check step; here, an agent deployment fails because the selected Azure region lacks model quota. During customer support center remediation, the proposed step to run sustained representative requests and verify no capacity or throttling failure is not the direct restoration action for regional capacity planning when an agent deployment fails because the selected Azure region lacks model quota; the keyed repair is request suitable capacity or use an approved available deployment region.

Question 89

A service administrator at a manufacturing group wants the following issue to be prevented in later deployments: An agent deployment fails because the selected Azure region lacks model quota. Which enduring control or engineering guardrail BEST prevents recurrence?

  1. Assume the same model quota and SKU exist in every Azure region
  2. Track regional quotas and expected peak request load
  3. Request suitable capacity or use an approved available deployment region
  4. Validate resource IDs and authorized project scopes in deployment pipelines

Explanation

Correct Answer

B:

The sustained prevention measure is to track regional quotas and expected peak request load. The implementation can still use the separate action to select a model deployment region and capacity allocation that support expected demand, but a guardrail constrains future changes. The manufacturing group can measure whether the rule works by arranging to run sustained representative requests and verify no capacity or throttling failure.

Incorrect Answers

A:

For prevention at the manufacturing group, the measure to assume the same model quota and SKU exist in every Azure region is weaker or aimed at another issue than the regional capacity planning guardrail to track regional quotas and expected peak request load needed when an agent deployment fails because the selected Azure region lacks model quota. The lasting guardrail for this specific threat is to track regional quotas and expected peak request load, not simply to complete an adjacent service task.

C:

The action to request suitable capacity or use an approved available deployment region can contribute to regional capacity planning as a corrective remediation step; here, an agent deployment fails because the selected Azure region lacks model quota. For prevention at the manufacturing group, the measure to request suitable capacity or use an approved available deployment region is weaker or aimed at another issue than the regional capacity planning guardrail to track regional quotas and expected peak request load needed when an agent deployment fails because the selected Azure region lacks model quota.

D:

The action to validate resource IDs and authorized project scopes in deployment pipelines addresses foundry endpoint routing rather than the decisive regional capacity planning condition. The manufacturing group should not infer that fixing one adjacent AI service validates its unrelated workflow; the requested action is to track regional quotas and expected peak request load.

Question 90

During a operational readiness review, a cloud developer must accept a change for regional capacity planning. The issue to be addressed is: An agent deployment fails because the selected Azure region lacks model quota. Which concrete verification action demonstrates that the relevant behavior works?

  1. Select a model deployment region and capacity allocation that support expected demand
  2. Assume the same model quota and SKU exist in every Azure region
  3. Run sustained representative requests and verify no capacity or throttling failure
  4. Send a harmless representative prompt and validate the model response and trace

Explanation

Correct Answer

C:

To verify this change, the legal operations unit should run sustained representative requests and verify no capacity or throttling failure. Operators still need the implementation to select a model deployment region and capacity allocation that support expected demand and the preventive requirement to track regional quotas and expected peak request load.

Incorrect Answers

A:

The action to select a model deployment region and capacity allocation that support expected demand can contribute to regional capacity planning as a implementation step; here, an agent deployment fails because the selected Azure region lacks model quota. In the legal operations unit acceptance review, trying to select a model deployment region and capacity allocation that support expected demand does not test the exact regional capacity planning condition where an agent deployment fails because the selected Azure region lacks model quota; the relevant test is to run sustained representative requests and verify no capacity or throttling failure.

B:

In the legal operations unit acceptance review, trying to assume the same model quota and SKU exist in every Azure region does not test the exact regional capacity planning condition where an agent deployment fails because the selected Azure region lacks model quota; the relevant test is to run sustained representative requests and verify no capacity or throttling failure. Acceptance should instead demonstrate that the service can run sustained representative requests and verify no capacity or throttling failure, using the described failure as the test criterion.

D:

In the legal operations unit acceptance review, trying to send a harmless representative prompt and validate the model response and trace does not test the exact regional capacity planning condition where an agent deployment fails because the selected Azure region lacks model quota; the relevant test is to run sustained representative requests and verify no capacity or throttling failure. Without the actual change to run sustained representative requests and verify no capacity or throttling failure, the answer lacks the correct operational scope even if the alternate action is technically valid elsewhere.

Developing AI Apps and Agents on Azure (AI-103) Practice Tests