McAfee-Secured Website

Exam Code: AI-300

Exam Name: Operationalizing Machine Learning and Generative AI Solutions

Certification Provider: Microsoft

Corresponding Certification: Microsoft Certified: Machine Learning Operations Engineer Associate

100% Microsoft AI-300 Guaranteed Success With Testking Training Materials

Get Real AI-300 Exam Questions & Expert Verified Answers as Experienced on the Real Exam!

213 Questions & Answers with Testing Engine

"Operationalizing Machine Learning and Generative AI Solutions Exam", also known as AI-300 exam, is a Microsoft certification exam.

Pass your tests with the always up-to-date AI-300 Exam Engine. Your AI-300 training materials keep you at the head of the pack!

Satisfaction Guaranteed

Money Back Guarantee

Testking provides hassle-free money back guarantee with our products. That is because we have 100% trust in the abilities of our professional and experience product team, and our record is a proof of that.

99.6% PASS RATE
Was: $137.49
Now: $124.99

Microsoft AI-300 Practice Test 30

Question 290

A hospital classification model has a high sensitivity score but very low precision, causing many unnecessary interventions. What should reviewers examine?

  1. Run representative missing-feature stress tests and evaluate resulting performance disparities.
  2. Evaluate error rates and fairness metrics separately for relevant demographic cohorts.
  3. Assess false-positive harms and the decision threshold under clinical constraints.
  4. Test proxy-feature impact and potentially reduce or constrain unfair reliance on that feature.

Explanation

Correct Answer

C:

A single score cannot capture competing error costs; deployment thresholds must reflect the harm of both error types.

Incorrect Answers

A:

Run representative missing-feature stress tests and evaluate resulting performance disparities. That operation would be relevant if: A data scientist wants to compare error patterns when one critical input feature is missing during inference. Its corresponding reason or verification is: Robustness checks reveal whether incomplete production records can trigger unsafe or systematically biased predictions. However the problem in this question is: A hospital classification model has a high sensitivity score but very low precision, causing many unnecessary interventions. The requirement instead calls for Assess false-positive harms and the decision threshold under clinical constraints; the question-specific check is: A single score cannot capture competing error costs; deployment thresholds must reflect the harm of both error types.

B:

Evaluate error rates and fairness metrics separately for relevant demographic cohorts. A separate source case supports that approach when: An automated lending classifier has high overall accuracy but significantly higher false-negative rates for one demographic group. The supporting rationale or observable result is: Aggregate performance can hide unequal outcomes, so group-specific analysis is needed before claiming responsible deployment. This question describes a different issue: A hospital classification model has a high sensitivity score but very low precision, causing many unnecessary interventions. Address it through Assess false-positive harms and the decision threshold under clinical constraints, with the required justification or evidence: A single score cannot capture competing error costs; deployment thresholds must reflect the harm of both error types.

D:

Test proxy-feature impact and potentially reduce or constrain unfair reliance on that feature. This is useful for the distinct issue: A responsible AI review finds the model relies heavily on a postcode proxy for protected characteristics. The relevant technical rationale or evidence for that issue is: High predictive utility does not excuse discriminatory proxy behavior; the feature's effects require explicit fairness analysis. Here the reported issue is: A hospital classification model has a high sensitivity score but very low precision, causing many unnecessary interventions. The relevant response is: Assess false-positive harms and the decision threshold under clinical constraints. The first result cannot verify this different correction: A single score cannot capture competing error costs; deployment thresholds must reflect the harm of both error types.

Question 291

A model performs well on common conditions but fails consistently on a small high-risk subgroup. What should the team do?

  1. Treat feature attribution as model explanation rather than definitive causal evidence.
  2. Expand targeted evaluation and mitigation for the underperforming subgroup before release.
  3. Assess false-positive harms and the decision threshold under clinical constraints.
  4. Run representative missing-feature stress tests and evaluate resulting performance disparities.

Explanation

Correct Answer

B:

Rare high-impact cohorts can be missed by average metrics; responsible release requires examining disproportionate failure risk.

Incorrect Answers

A:

The alternative, Treat feature attribution as model explanation rather than definitive causal evidence, belongs to a different technical condition: An explainability report claims feature importance proves a causal relationship between income and outcomes. There the source-specific mechanism or proof is: Predictive associations may arise from correlated variables; explanation weights alone do not establish cause and effect. The condition asked about here is: A model performs well on common conditions but fails consistently on a small high-risk subgroup. A suitable correction is Expand targeted evaluation and mitigation for the underperforming subgroup before release; the targeted justification or result is: Rare high-impact cohorts can be missed by average metrics; responsible release requires examining disproportionate failure risk.

C:

Assess false-positive harms and the decision threshold under clinical constraints. That operation would be relevant if: A hospital classification model has a high sensitivity score but very low precision, causing many unnecessary interventions. Its corresponding reason or verification is: A single score cannot capture competing error costs; deployment thresholds must reflect the harm of both error types. However the problem in this question is: A model performs well on common conditions but fails consistently on a small high-risk subgroup. The requirement instead calls for Expand targeted evaluation and mitigation for the underperforming subgroup before release; the question-specific check is: Rare high-impact cohorts can be missed by average metrics; responsible release requires examining disproportionate failure risk.

D:

Run representative missing-feature stress tests and evaluate resulting performance disparities. A separate source case supports that approach when: A data scientist wants to compare error patterns when one critical input feature is missing during inference. The supporting rationale or observable result is: Robustness checks reveal whether incomplete production records can trigger unsafe or systematically biased predictions. This question describes a different issue: A model performs well on common conditions but fails consistently on a small high-risk subgroup. Address it through Expand targeted evaluation and mitigation for the underperforming subgroup before release, with the required justification or evidence: Rare high-impact cohorts can be missed by average metrics; responsible release requires examining disproportionate failure risk.

Question 292

An explainability report claims feature importance proves a causal relationship between income and outcomes. What interpretation should be corrected?

  1. Treat feature attribution as model explanation rather than definitive causal evidence.
  2. Review data provenance, consent, and permitted feature usage alongside performance evidence.
  3. Use an approved privacy-preserving explainability workflow with appropriate access controls.
  4. Evaluate error rates and fairness metrics separately for relevant demographic cohorts.

Explanation

Correct Answer

A:

Predictive associations may arise from correlated variables; explanation weights alone do not establish cause and effect.

Incorrect Answers

B:

Review data provenance, consent, and permitted feature usage alongside performance evidence. It addresses another case: A responsible AI review considers a model's prediction accuracy but ignores whether its inputs are permissible for the intended purpose. For that case, the distinguishing justification or acceptance check is: Responsible deployment depends on lawful and appropriate data use in addition to technical predictive performance. The present case instead states: An explainability report claims feature importance proves a causal relationship between income and outcomes. It calls for Treat feature attribution as model explanation rather than definitive causal evidence. Showing success for the other case would not establish: Predictive associations may arise from correlated variables; explanation weights alone do not establish cause and effect.

C:

The alternative, Use an approved privacy-preserving explainability workflow with appropriate access controls, belongs to a different technical condition: An engineer sends raw personal identifiers to an unapproved external explainability service. There the source-specific mechanism or proof is: Model interpretation must still respect personal-data minimization and authorized processing requirements for sensitive records. The condition asked about here is: An explainability report claims feature importance proves a causal relationship between income and outcomes. A suitable correction is Treat feature attribution as model explanation rather than definitive causal evidence; the targeted justification or result is: Predictive associations may arise from correlated variables; explanation weights alone do not establish cause and effect.

D:

Evaluate error rates and fairness metrics separately for relevant demographic cohorts. That operation would be relevant if: An automated lending classifier has high overall accuracy but significantly higher false-negative rates for one demographic group. Its corresponding reason or verification is: Aggregate performance can hide unequal outcomes, so group-specific analysis is needed before claiming responsible deployment. However the problem in this question is: An explainability report claims feature importance proves a causal relationship between income and outcomes. The requirement instead calls for Treat feature attribution as model explanation rather than definitive causal evidence; the question-specific check is: Predictive associations may arise from correlated variables; explanation weights alone do not establish cause and effect.

Question 293

A data scientist wants to compare error patterns when one critical input feature is missing during inference. What should the evaluation include?

  1. Treat feature attribution as model explanation rather than definitive causal evidence.
  2. Use an approved privacy-preserving explainability workflow with appropriate access controls.
  3. Expand targeted evaluation and mitigation for the underperforming subgroup before release.
  4. Run representative missing-feature stress tests and evaluate resulting performance disparities.

Explanation

Correct Answer

D:

Robustness checks reveal whether incomplete production records can trigger unsafe or systematically biased predictions.

Incorrect Answers

A:

Treat feature attribution as model explanation rather than definitive causal evidence. This is useful for the distinct issue: An explainability report claims feature importance proves a causal relationship between income and outcomes. The relevant technical rationale or evidence for that issue is: Predictive associations may arise from correlated variables; explanation weights alone do not establish cause and effect. Here the reported issue is: A data scientist wants to compare error patterns when one critical input feature is missing during inference. The relevant response is: Run representative missing-feature stress tests and evaluate resulting performance disparities. The first result cannot verify this different correction: Robustness checks reveal whether incomplete production records can trigger unsafe or systematically biased predictions.

B:

Use an approved privacy-preserving explainability workflow with appropriate access controls. It addresses another case: An engineer sends raw personal identifiers to an unapproved external explainability service. For that case, the distinguishing justification or acceptance check is: Model interpretation must still respect personal-data minimization and authorized processing requirements for sensitive records. The present case instead states: A data scientist wants to compare error patterns when one critical input feature is missing during inference. It calls for Run representative missing-feature stress tests and evaluate resulting performance disparities. Showing success for the other case would not establish: Robustness checks reveal whether incomplete production records can trigger unsafe or systematically biased predictions.

C:

The alternative, Expand targeted evaluation and mitigation for the underperforming subgroup before release, belongs to a different technical condition: A model performs well on common conditions but fails consistently on a small high-risk subgroup. There the source-specific mechanism or proof is: Rare high-impact cohorts can be missed by average metrics; responsible release requires examining disproportionate failure risk. The condition asked about here is: A data scientist wants to compare error patterns when one critical input feature is missing during inference. A suitable correction is Run representative missing-feature stress tests and evaluate resulting performance disparities; the targeted justification or result is: Robustness checks reveal whether incomplete production records can trigger unsafe or systematically biased predictions.

Question 294

A model owner requests a fairness conclusion based on only three examples from the smallest cohort. What is the correct review response?

  1. Document known limitations, risk tradeoffs, and intended-use constraints in the model review.
  2. Produce validated feature-attribution or local explainability analyses for representative cases.
  3. Collect sufficient cohort evidence or document the uncertainty before approving fairness claims.
  4. Evaluate representative cohorts and distribution shifts across the intended deployment populations.

Explanation

Correct Answer

C:

Tiny samples can make subgroup metrics unstable, so conclusions should reflect their uncertainty and limited statistical power.

Incorrect Answers

A:

Document known limitations, risk tradeoffs, and intended-use constraints in the model review. A separate source case supports that approach when: A regulated classifier passes accuracy tests but its failure cases are not documented for downstream operational reviewers. The supporting rationale or observable result is: Responsible AI governance includes contextual limitations and misuse risks, not only a numerical performance score. This question describes a different issue: A model owner requests a fairness conclusion based on only three examples from the smallest cohort. Address it through Collect sufficient cohort evidence or document the uncertainty before approving fairness claims, with the required justification or evidence: Tiny samples can make subgroup metrics unstable, so conclusions should reflect their uncertainty and limited statistical power.

B:

Produce validated feature-attribution or local explainability analyses for representative cases. This is useful for the distinct issue: A clinician wants to understand which input features drove individual predictions from a model used for triage. The relevant technical rationale or evidence for that issue is: Local explanations help reviewers examine model decision factors rather than assuming overall accuracy explains each outcome. Here the reported issue is: A model owner requests a fairness conclusion based on only three examples from the smallest cohort. The relevant response is: Collect sufficient cohort evidence or document the uncertainty before approving fairness claims. The first result cannot verify this different correction: Tiny samples can make subgroup metrics unstable, so conclusions should reflect their uncertainty and limited statistical power.

D:

Evaluate representative cohorts and distribution shifts across the intended deployment populations. It addresses another case: A safety review uses only training data from one country although the model will be used internationally. For that case, the distinguishing justification or acceptance check is: Responsible generalization testing requires evidence from the people and contexts the model is expected to serve. The present case instead states: A model owner requests a fairness conclusion based on only three examples from the smallest cohort. It calls for Collect sufficient cohort evidence or document the uncertainty before approving fairness claims. Showing success for the other case would not establish: Tiny samples can make subgroup metrics unstable, so conclusions should reflect their uncertainty and limited statistical power.

Question 295

A responsible AI review considers a model's prediction accuracy but ignores whether its inputs are permissible for the intended purpose. What should be added?

  1. Evaluate representative cohorts and distribution shifts across the intended deployment populations.
  2. Review data provenance, consent, and permitted feature usage alongside performance evidence.
  3. Test proxy-feature impact and potentially reduce or constrain unfair reliance on that feature.
  4. Expand targeted evaluation and mitigation for the underperforming subgroup before release.

Explanation

Correct Answer

B:

Responsible deployment depends on lawful and appropriate data use in addition to technical predictive performance.

Incorrect Answers

A:

Evaluate representative cohorts and distribution shifts across the intended deployment populations. That operation would be relevant if: A safety review uses only training data from one country although the model will be used internationally. Its corresponding reason or verification is: Responsible generalization testing requires evidence from the people and contexts the model is expected to serve. However the problem in this question is: A responsible AI review considers a model's prediction accuracy but ignores whether its inputs are permissible for the intended purpose. The requirement instead calls for Review data provenance, consent, and permitted feature usage alongside performance evidence; the question-specific check is: Responsible deployment depends on lawful and appropriate data use in addition to technical predictive performance.

C:

Test proxy-feature impact and potentially reduce or constrain unfair reliance on that feature. A separate source case supports that approach when: A responsible AI review finds the model relies heavily on a postcode proxy for protected characteristics. The supporting rationale or observable result is: High predictive utility does not excuse discriminatory proxy behavior; the feature's effects require explicit fairness analysis. This question describes a different issue: A responsible AI review considers a model's prediction accuracy but ignores whether its inputs are permissible for the intended purpose. Address it through Review data provenance, consent, and permitted feature usage alongside performance evidence, with the required justification or evidence: Responsible deployment depends on lawful and appropriate data use in addition to technical predictive performance.

D:

Expand targeted evaluation and mitigation for the underperforming subgroup before release. This is useful for the distinct issue: A model performs well on common conditions but fails consistently on a small high-risk subgroup. The relevant technical rationale or evidence for that issue is: Rare high-impact cohorts can be missed by average metrics; responsible release requires examining disproportionate failure risk. Here the reported issue is: A responsible AI review considers a model's prediction accuracy but ignores whether its inputs are permissible for the intended purpose. The relevant response is: Review data provenance, consent, and permitted feature usage alongside performance evidence. The first result cannot verify this different correction: Responsible deployment depends on lawful and appropriate data use in addition to technical predictive performance.

Question 296

A registry contains an old model still used by one critical production endpoint. An engineer proposes deleting every deprecated version. What should happen first?

  1. Inventory active deployment dependencies before retiring or deleting model versions.
  2. Record the deprecation reason, reviewer decision, and replacement version alongside lifecycle state.
  3. Use a stable registered model name and explicit version rather than transient storage paths.
  4. Archive or remove obsolete model versions only after dependency and retention checks.

Explanation

Correct Answer

A:

A model version must not be removed while serving workloads still depend on its artifact and lineage.

Incorrect Answers

B:

The alternative, Record the deprecation reason, reviewer decision, and replacement version alongside lifecycle state, belongs to a different technical condition: A team needs to show why an older model version stopped serving due to fairness concerns. There the source-specific mechanism or proof is: Governance metadata explains retirement decisions and prevents unsafe unreviewed reactivation of an old artifact. The condition asked about here is: A registry contains an old model still used by one critical production endpoint. An engineer proposes deleting every deprecated version. A suitable correction is Inventory active deployment dependencies before retiring or deleting model versions; the targeted justification or result is: A model version must not be removed while serving workloads still depend on its artifact and lineage.

C:

Use a stable registered model name and explicit version rather than transient storage paths. That operation would be relevant if: A production rollout references a model by its environment-specific storage URI, which changes after a workspace migration. Its corresponding reason or verification is: Registry identifiers separate logical model identity from storage location, improving portability across controlled deployments. However the problem in this question is: A registry contains an old model still used by one critical production endpoint. An engineer proposes deleting every deprecated version. The requirement instead calls for Inventory active deployment dependencies before retiring or deleting model versions; the question-specific check is: A model version must not be removed while serving workloads still depend on its artifact and lineage.

D:

Archive or remove obsolete model versions only after dependency and retention checks. A separate source case supports that approach when: A decommissioned application no longer uses a production model, and storage costs grow with obsolete artifacts. The supporting rationale or observable result is: Safe lifecycle cleanup balances operational cost with rollback needs and legally required artifact retention. This question describes a different issue: A registry contains an old model still used by one critical production endpoint. An engineer proposes deleting every deprecated version. Address it through Inventory active deployment dependencies before retiring or deleting model versions, with the required justification or evidence: A model version must not be removed while serving workloads still depend on its artifact and lineage.

Question 297

A new candidate passes testing, but operators want a way to restore last week's production model quickly. What lifecycle policy supports this?

  1. Promote the immutable historical version through a new documented deployment action.
  2. Record immutable review outcomes and approval timestamps with the model release metadata.
  3. Use explicit model promotion approvals and version pinning in release manifests.
  4. Retain an approved previous model version for controlled rollback.

Explanation

Correct Answer

D:

Preserving prior deployable artifacts supports recovery when a new model fails under real traffic.

Incorrect Answers

A:

Promote the immutable historical version through a new documented deployment action. It addresses another case: An operator tries to roll back by editing an older model artifact in place rather than deploying its recorded version. For that case, the distinguishing justification or acceptance check is: Rollback should restore a known artifact while preserving both the original failure and corrective deployment history. The present case instead states: A new candidate passes testing, but operators want a way to restore last week's production model quickly. It calls for Retain an approved previous model version for controlled rollback. Showing success for the other case would not establish: Preserving prior deployable artifacts supports recovery when a new model fails under real traffic.

B:

The alternative, Record immutable review outcomes and approval timestamps with the model release metadata, belongs to a different technical condition: An organization wants approval history for model versions, but reviewers communicate decisions only in private chat. There the source-specific mechanism or proof is: Audit trails should attach formal decisions to exact versions rather than relying on scattered conversations. The condition asked about here is: A new candidate passes testing, but operators want a way to restore last week's production model quickly. A suitable correction is Retain an approved previous model version for controlled rollback; the targeted justification or result is: Preserving prior deployable artifacts supports recovery when a new model fails under real traffic.

C:

Use explicit model promotion approvals and version pinning in release manifests. That operation would be relevant if: An experimental registry has dozens of unapproved models; a deployment script selects newest automatically. Its corresponding reason or verification is: A new registry entry is not necessarily production ready; controlled promotion distinguishes approval from chronological creation. However the problem in this question is: A new candidate passes testing, but operators want a way to restore last week's production model quickly. The requirement instead calls for Retain an approved previous model version for controlled rollback; the question-specific check is: Preserving prior deployable artifacts supports recovery when a new model fails under real traffic.

Question 298

An experimental registry has dozens of unapproved models; a deployment script selects newest automatically. What governance change is needed?

  1. Promote the immutable historical version through a new documented deployment action.
  2. Record immutable review outcomes and approval timestamps with the model release metadata.
  3. Use explicit model promotion approvals and version pinning in release manifests.
  4. Use a stable registered model name and explicit version rather than transient storage paths.

Explanation

Correct Answer

C:

A new registry entry is not necessarily production ready; controlled promotion distinguishes approval from chronological creation.

Incorrect Answers

A:

Promote the immutable historical version through a new documented deployment action. This is useful for the distinct issue: An operator tries to roll back by editing an older model artifact in place rather than deploying its recorded version. The relevant technical rationale or evidence for that issue is: Rollback should restore a known artifact while preserving both the original failure and corrective deployment history. Here the reported issue is: An experimental registry has dozens of unapproved models; a deployment script selects newest automatically. The relevant response is: Use explicit model promotion approvals and version pinning in release manifests. The first result cannot verify this different correction: A new registry entry is not necessarily production ready; controlled promotion distinguishes approval from chronological creation.

B:

Record immutable review outcomes and approval timestamps with the model release metadata. It addresses another case: An organization wants approval history for model versions, but reviewers communicate decisions only in private chat. For that case, the distinguishing justification or acceptance check is: Audit trails should attach formal decisions to exact versions rather than relying on scattered conversations. The present case instead states: An experimental registry has dozens of unapproved models; a deployment script selects newest automatically. It calls for Use explicit model promotion approvals and version pinning in release manifests. Showing success for the other case would not establish: A new registry entry is not necessarily production ready; controlled promotion distinguishes approval from chronological creation.

D:

The alternative, Use a stable registered model name and explicit version rather than transient storage paths, belongs to a different technical condition: A production rollout references a model by its environment-specific storage URI, which changes after a workspace migration. There the source-specific mechanism or proof is: Registry identifiers separate logical model identity from storage location, improving portability across controlled deployments. The condition asked about here is: An experimental registry has dozens of unapproved models; a deployment script selects newest automatically. A suitable correction is Use explicit model promotion approvals and version pinning in release manifests; the targeted justification or result is: A new registry entry is not necessarily production ready; controlled promotion distinguishes approval from chronological creation.

Question 299

A compliance audit asks which dataset and code produced a model version now retired from serving. What should remain accessible?

  1. Record immutable review outcomes and approval timestamps with the model release metadata.
  2. Retain the producing run, data versions, and code/environment lineage with the archived model.
  3. Archive or remove obsolete model versions only after dependency and retention checks.
  4. Use a stable registered model name and explicit version rather than transient storage paths.

Explanation

Correct Answer

B:

Archiving should preserve the evidence needed to reconstruct the model's origin despite removing it from active serving.

Incorrect Answers

A:

Record immutable review outcomes and approval timestamps with the model release metadata. A separate source case supports that approach when: An organization wants approval history for model versions, but reviewers communicate decisions only in private chat. The supporting rationale or observable result is: Audit trails should attach formal decisions to exact versions rather than relying on scattered conversations. This question describes a different issue: A compliance audit asks which dataset and code produced a model version now retired from serving. Address it through Retain the producing run, data versions, and code/environment lineage with the archived model, with the required justification or evidence: Archiving should preserve the evidence needed to reconstruct the model's origin despite removing it from active serving.

C:

Archive or remove obsolete model versions only after dependency and retention checks. This is useful for the distinct issue: A decommissioned application no longer uses a production model, and storage costs grow with obsolete artifacts. The relevant technical rationale or evidence for that issue is: Safe lifecycle cleanup balances operational cost with rollback needs and legally required artifact retention. Here the reported issue is: A compliance audit asks which dataset and code produced a model version now retired from serving. The relevant response is: Retain the producing run, data versions, and code/environment lineage with the archived model. The first result cannot verify this different correction: Archiving should preserve the evidence needed to reconstruct the model's origin despite removing it from active serving.

D:

Use a stable registered model name and explicit version rather than transient storage paths. It addresses another case: A production rollout references a model by its environment-specific storage URI, which changes after a workspace migration. For that case, the distinguishing justification or acceptance check is: Registry identifiers separate logical model identity from storage location, improving portability across controlled deployments. The present case instead states: A compliance audit asks which dataset and code produced a model version now retired from serving. It calls for Retain the producing run, data versions, and code/environment lineage with the archived model. Showing success for the other case would not establish: Archiving should preserve the evidence needed to reconstruct the model's origin despite removing it from active serving.

Question 300

A decommissioned application no longer uses a production model, and storage costs grow with obsolete artifacts. What cleanup strategy is appropriate?

  1. Archive or remove obsolete model versions only after dependency and retention checks.
  2. Maintain accountable owner and lifecycle status metadata for every registered model family.
  3. Record the deprecation reason, reviewer decision, and replacement version alongside lifecycle state.
  4. Use a stable registered model name and explicit version rather than transient storage paths.

Explanation

Correct Answer

A:

Safe lifecycle cleanup balances operational cost with rollback needs and legally required artifact retention.

Incorrect Answers

B:

Maintain accountable owner and lifecycle status metadata for every registered model family. This is useful for the distinct issue: A data science team has no record of who owns models abandoned in a shared registry. The relevant technical rationale or evidence for that issue is: Ownership information enables safe review, maintenance and retirement decisions when model-producing staff or projects change. Here the reported issue is: A decommissioned application no longer uses a production model, and storage costs grow with obsolete artifacts. The relevant response is: Archive or remove obsolete model versions only after dependency and retention checks. The first result cannot verify this different correction: Safe lifecycle cleanup balances operational cost with rollback needs and legally required artifact retention.

C:

Record the deprecation reason, reviewer decision, and replacement version alongside lifecycle state. It addresses another case: A team needs to show why an older model version stopped serving due to fairness concerns. For that case, the distinguishing justification or acceptance check is: Governance metadata explains retirement decisions and prevents unsafe unreviewed reactivation of an old artifact. The present case instead states: A decommissioned application no longer uses a production model, and storage costs grow with obsolete artifacts. It calls for Archive or remove obsolete model versions only after dependency and retention checks. Showing success for the other case would not establish: Safe lifecycle cleanup balances operational cost with rollback needs and legally required artifact retention.

D:

The alternative, Use a stable registered model name and explicit version rather than transient storage paths, belongs to a different technical condition: A production rollout references a model by its environment-specific storage URI, which changes after a workspace migration. There the source-specific mechanism or proof is: Registry identifiers separate logical model identity from storage location, improving portability across controlled deployments. The condition asked about here is: A decommissioned application no longer uses a production model, and storage costs grow with obsolete artifacts. A suitable correction is Archive or remove obsolete model versions only after dependency and retention checks; the targeted justification or result is: Safe lifecycle cleanup balances operational cost with rollback needs and legally required artifact retention.

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