Microsoft AI-300 (Operationalizing Machine Learning and Generative AI Solutions) Practice Test 31

Question 300

An inference client fails after an app is moved to a different Foundry project because the developer reused an endpoint and deployment alias from the old project. Which project metadata should the team discover before updating the inference client configuration?

  1. Project teardown plan
  2. Foundry endpoint discovery
  3. Project-level ownership
  4. Connection inventory

Explanation

Correct Answer

B:

Applications must call the correct endpoint and model or deployment identifier. A project display title is not interchangeable with the serving API target.

Incorrect Answers

A:

Decommissioning reduces abandoned resources and exposure; evidence retention and shared dependencies must be considered before deleting a project; the outstanding need is to determine the supported endpoint and deployment name to configure an application's inference client.

C:

Clearly defined ownership separates development from governance; cloud role scope still determines which management actions each maintainer can perform; the needed capability is to determine the supported endpoint and deployment name to configure an application's inference client.

D:

An inventory reveals dependencies such as search, storage, and model endpoints; copying prompt text without its connections may leave the application unable to run; the requirement, by contrast, is to determine the supported endpoint and deployment name to configure an application's inference client.

Question 301

Security reviewers require secretless service-to-service access for generative AI applications. The immediate design goal is to authenticate a supported Azure-hosted application to its model service without embedding API keys in source code. Which approach addresses that goal?

  1. Service principal credential lifecycle
  2. Token audience selection
  3. Identity-specific audit logs
  4. Role-based access control
  5. Managed identity for application

Explanation

Correct Answer

E:

Managed identity obtains tokens for the running Azure resource. The identity still needs suitable authorization on the target inference service.

Incorrect Answers

A:

A service principal may authenticate using supported credentials, which require secure storage and rotation; it should not be confused with a human's interactive account; the requirement, by contrast, is to authenticate a supported Azure-hosted application to its model service without embedding API keys in source code.

B:

Tokens are issued for a particular target audience or resource; an otherwise valid token may be rejected by a service that expects another audience; the team would still need to authenticate a supported Azure-hosted application to its model service without embedding API keys in source code.

C:

Identity-aware logs support attribution of administrative and service calls; shared keys make individual actor attribution harder without complementary controls; the specified outcome instead requires the team to authenticate a supported Azure-hosted application to its model service without embedding API keys in source code.

D:

RBAC determines permitted actions for identities after authentication; a valid identity token alone does not grant every Foundry management operation; the implementation under review needs to authenticate a supported Azure-hosted application to its model service without embedding API keys in source code.

Question 302

Security reviewers require secretless service-to-service access for generative AI applications. The team is required to assign only the actions needed for model deployment or consumption at the appropriate Azure scope. Which option is the best fit?

  1. Role-based access control
  2. Service principal credential lifecycle
  3. Token audience selection
  4. Managed identity for application

Explanation

Correct Answer

A:

RBAC determines permitted actions for identities after authentication. A valid identity token alone does not grant every Foundry management operation.

Incorrect Answers

B:

A service principal may authenticate using supported credentials, which require secure storage and rotation; it should not be confused with a human's interactive account; the requested task remains to assign only the actions needed for model deployment or consumption at the appropriate Azure scope.

C:

Tokens are issued for a particular target audience or resource; an otherwise valid token may be rejected by a service that expects another audience; this alternative does not achieve the requirement to assign only the actions needed for model deployment or consumption at the appropriate Azure scope.

D:

Managed identity obtains tokens for the running Azure resource; the identity still needs suitable authorization on the target inference service; the actual design criterion is to assign only the actions needed for model deployment or consumption at the appropriate Azure scope.

Question 303

Security reviewers require secretless service-to-service access for generative AI applications. An operator is tasked to share one deliberately managed identity across eligible app services without binding it to one resource lifecycle. Which mechanism provides that capability?

  1. User-assigned workload identity
  2. Role-based access control
  3. Identity-specific audit logs
  4. Token audience selection

Explanation

Correct Answer

A:

User-assigned identities allow separate lifecycle and access management. Their permissions apply wherever the identity is attached and authenticated.

Incorrect Answers

B:

RBAC determines permitted actions for identities after authentication; a valid identity token alone does not grant every Foundry management operation; the outstanding need is to share one deliberately managed identity across eligible app services without binding it to one resource lifecycle.

C:

Identity-aware logs support attribution of administrative and service calls; shared keys make individual actor attribution harder without complementary controls; this approach leaves the team unable to share one deliberately managed identity across eligible app services without binding it to one resource lifecycle.

D:

Tokens are issued for a particular target audience or resource; an otherwise valid token may be rejected by a service that expects another audience; the needed capability is to share one deliberately managed identity across eligible app services without binding it to one resource lifecycle.

Question 304

Security reviewers require secretless service-to-service access for generative AI applications. The immediate design goal is to rotate a legacy application credential when federation or managed identity cannot be used for that integration. Which approach addresses that goal?

  1. Managed identity for application
  2. Identity-specific audit logs
  3. Service principal credential lifecycle
  4. Token audience selection

Explanation

Correct Answer

C:

A service principal may authenticate using supported credentials, which require secure storage and rotation. It should not be confused with a human's interactive account.

Incorrect Answers

A:

Managed identity obtains tokens for the running Azure resource; the identity still needs suitable authorization on the target inference service; the needed capability is to rotate a legacy application credential when federation or managed identity cannot be used for that integration.

B:

Identity-aware logs support attribution of administrative and service calls; shared keys make individual actor attribution harder without complementary controls; the requirement, by contrast, is to rotate a legacy application credential when federation or managed identity cannot be used for that integration.

D:

Tokens are issued for a particular target audience or resource; an otherwise valid token may be rejected by a service that expects another audience; the specified outcome instead requires the team to rotate a legacy application credential when federation or managed identity cannot be used for that integration.

Question 305

Security reviewers require secretless service-to-service access for generative AI applications. The team is required to diagnose an authenticated call that fails because its access token targets the wrong resource audience. Which option is the best fit?

  1. Managed identity for application
  2. Token audience selection
  3. Identity-specific audit logs
  4. User-assigned workload identity

Explanation

Correct Answer

B:

Tokens are issued for a particular target audience or resource. An otherwise valid token may be rejected by a service that expects another audience.

Incorrect Answers

A:

Managed identity obtains tokens for the running Azure resource; the identity still needs suitable authorization on the target inference service; the specified outcome instead requires the team to diagnose an authenticated call that fails because its access token targets the wrong resource audience.

C:

Identity-aware logs support attribution of administrative and service calls; shared keys make individual actor attribution harder without complementary controls; the requested task remains to diagnose an authenticated call that fails because its access token targets the wrong resource audience.

D:

User-assigned identities allow separate lifecycle and access management; their permissions apply wherever the identity is attached and authenticated; this alternative does not achieve the requirement to diagnose an authenticated call that fails because its access token targets the wrong resource audience.

Question 306

Security reviewers require secretless service-to-service access for generative AI applications. An operator is tasked to determine which nonhuman identity initiated a model deployment or accessed a restricted project resource. Which mechanism provides that capability?

  1. Role-based access control
  2. Managed identity for application
  3. Identity-specific audit logs
  4. User-assigned workload identity

Explanation

Correct Answer

C:

Identity-aware logs support attribution of administrative and service calls. Shared keys make individual actor attribution harder without complementary controls.

Incorrect Answers

A:

RBAC determines permitted actions for identities after authentication; a valid identity token alone does not grant every Foundry management operation; this alternative does not achieve the requirement to determine which nonhuman identity initiated a model deployment or accessed a restricted project resource.

B:

Managed identity obtains tokens for the running Azure resource; the identity still needs suitable authorization on the target inference service; the actual design criterion is to determine which nonhuman identity initiated a model deployment or accessed a restricted project resource.

D:

User-assigned identities allow separate lifecycle and access management; their permissions apply wherever the identity is attached and authenticated; this approach leaves the team unable to determine which nonhuman identity initiated a model deployment or accessed a restricted project resource.

Question 307

A Foundry operations group separates infrastructure administrators from application developers. The immediate design goal is to allow an operator to configure model resource deployments under an authorized management scope. Which approach addresses that goal?

  1. Least-privilege access test
  2. Resource management permission
  3. Privileged role assignment review
  4. Permission propagation check

Explanation

Correct Answer

B:

Management-plane roles govern resource configuration actions. They should be distinguished from rights to submit individual inference requests.

Incorrect Answers

A:

Positive and negative tests confirm the intended permission boundary; successful inference alone cannot prove that administrative access is sufficiently restricted; this approach leaves the team unable to allow an operator to configure model resource deployments under an authorized management scope.

C:

Role-assignment changes are sensitive authorization operations; standard resource editing privileges should not be assumed sufficient to delegate broad access; the requirement, by contrast, is to allow an operator to configure model resource deployments under an authorized management scope.

D:

Authorization changes can take time to become effective across services; retrying without checking token freshness and scope can misdiagnose a transient state; the team would still need to allow an operator to configure model resource deployments under an authorized management scope.

Question 308

A Foundry operations group separates infrastructure administrators from application developers. The team is required to permit an application to call an approved deployed model without granting permission to modify infrastructure. Which option is the best fit?

  1. Least-privilege access test
  2. Resource management permission
  3. Permission propagation check
  4. Inference data-plane access
  5. Project access restriction

Explanation

Correct Answer

D:

Data-plane access authorizes supported inference operations. It does not imply the caller can redeploy the underlying model.

Incorrect Answers

A:

Positive and negative tests confirm the intended permission boundary; successful inference alone cannot prove that administrative access is sufficiently restricted; the team would still need to permit an application to call an approved deployed model without granting permission to modify infrastructure.

B:

Management-plane roles govern resource configuration actions; they should be distinguished from rights to submit individual inference requests; the specified outcome instead requires the team to permit an application to call an approved deployed model without granting permission to modify infrastructure.

C:

Authorization changes can take time to become effective across services; retrying without checking token freshness and scope can misdiagnose a transient state; the implementation under review needs to permit an application to call an approved deployed model without granting permission to modify infrastructure.

E:

Narrow scope reduces exposure to other workloads; a broad subscription assignment can unintentionally grant visibility or management across projects; this alternative does not achieve the requirement to permit an application to call an approved deployed model without granting permission to modify infrastructure.

Question 309

A Foundry operations group separates infrastructure administrators from application developers. An operator is tasked to confine a contractor's collaboration rights to one approved Foundry project rather than a whole subscription. Which mechanism provides that capability?

  1. Permission propagation check
  2. Resource management permission
  3. Project access restriction
  4. Privileged role assignment review

Explanation

Correct Answer

C:

Narrow scope reduces exposure to other workloads. A broad subscription assignment can unintentionally grant visibility or management across projects.

Incorrect Answers

A:

Authorization changes can take time to become effective across services; retrying without checking token freshness and scope can misdiagnose a transient state; the requested task remains to confine a contractor's collaboration rights to one approved Foundry project rather than a whole subscription.

B:

Management-plane roles govern resource configuration actions; they should be distinguished from rights to submit individual inference requests; this alternative does not achieve the requirement to confine a contractor's collaboration rights to one approved Foundry project rather than a whole subscription.

D:

Role-assignment changes are sensitive authorization operations; standard resource editing privileges should not be assumed sufficient to delegate broad access; the outstanding need is to confine a contractor's collaboration rights to one approved Foundry project rather than a whole subscription.

Question 310

A Foundry operations group separates infrastructure administrators from application developers. The immediate design goal is to require an authorized access administrator when changing who may grant project or resource roles. Which approach addresses that goal?

  1. Inference data-plane access
  2. Resource management permission
  3. Project access restriction
  4. Privileged role assignment review

Explanation

Correct Answer

D:

Role-assignment changes are sensitive authorization operations. Standard resource editing privileges should not be assumed sufficient to delegate broad access.

Incorrect Answers

A:

Data-plane access authorizes supported inference operations; it does not imply the caller can redeploy the underlying model; the outstanding need is to require an authorized access administrator when changing who may grant project or resource roles.

B:

Management-plane roles govern resource configuration actions; they should be distinguished from rights to submit individual inference requests; this approach leaves the team unable to require an authorized access administrator when changing who may grant project or resource roles.

C:

Narrow scope reduces exposure to other workloads; a broad subscription assignment can unintentionally grant visibility or management across projects; the needed capability is to require an authorized access administrator when changing who may grant project or resource roles.

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