McAfee Secure

Exam Code: AI-901

Exam Name: Microsoft Azure AI Fundamentals

Certification Provider: Microsoft

Corresponding Certification: Microsoft Certified: Azure AI Fundamentals

Microsoft AI-901 Questions & Answers

Study with Up-To-Date REAL Exam Questions and Answers from the ACTUAL Test

166 Questions & Answers with Testing Engine

"Microsoft Azure AI Fundamentals Exam", also known as AI-901 exam, is a Microsoft certification exam.

Pass your tests with the always up-to-date AI-901 Exam Engine. Your AI-901 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-901 Practice Test 51

Question 500

A solution team has selected this control: Use a video-capable Content Understanding analyzer with timestamps and supported fields. Which observation is the most direct acceptance evidence?

  1. The answer resolves to a specific approved source document version
  2. The extracted steps map to the correct clips and source times
  3. Quoted text matches the time-aligned source transcript
  4. Observed actions link to supported source intervals while unobserved short events are not claimed as detected

Explanation

Correct Answer

B:

Multimodal extraction draws on visual and spoken evidence over time. The expected acceptance evidence is: The extracted steps map to the correct clips and source times. This observation directly tests the stated behavior.

Incorrect Answers

A:

Verifiable provenance differentiates a cited answer from guesswork. This observation does not establish the required result: The extracted steps map to the correct clips and source times.

C:

Generated summaries are paraphrases and not guaranteed verbatim quotes. This observation does not establish the required result: The extracted steps map to the correct clips and source times.

D:

Content Understanding samples approximately one frame per second, so brief or single-frame actions can be missed despite valid timestamps. This observation does not establish the required result: The extracted steps map to the correct clips and source times.

Question 501

A project needs the following AI behavior: A media processor must extract actions from a security-camera video. What should the designer implement?

  1. Distinguish transcript passages from generated summaries and retain timestamps
  2. Use timestamped multi-segment video analysis for visible actions, with separate higher-frame-rate review when brief security events matter
  3. Extract checkbox selection marks with supported document analysis, and verify the visible signature separately; use signature detection only if its preview API is approved
  4. Use audio extraction followed by grounded action-item analysis

Explanation

Correct Answer

B:

Content Understanding samples approximately one frame per second, so brief or single-frame actions can be missed despite valid timestamps.

Incorrect Answers

A:

Generated summaries are paraphrases and not guaranteed verbatim quotes. That is not sufficient here: Content Understanding samples approximately one frame per second, so brief or single-frame actions can be missed despite valid timestamps.

C:

Selection-mark extraction is available with GA document analysis, but dedicated signature-location detection is available only in the 2026-06-01-preview API and cannot establish legal validity. That is not sufficient here: Content Understanding samples approximately one frame per second, so brief or single-frame actions can be missed despite valid timestamps.

D:

Transcription and summarization are separate results with different validation. That is not sufficient here: Content Understanding samples approximately one frame per second, so brief or single-frame actions can be missed despite valid timestamps.

Question 502

Troubleshooting report: The model gives unverified treatment advice without escalation. Which correction is MOST directly targeted to this failure?

  1. Propagate typed tool errors and use an explicit unavailable response
  2. Raise high-risk routing and restrict unsupported actionable advice
  3. Gate the payment tool with verified role and amount checks
  4. Partition state by user and clear or secure session context

Explanation

Correct Answer

B:

This correction directly addresses the reported failure. Safety mechanisms should reduce unsupported high-impact recommendations. The targeted change is: Raise high-risk routing and restrict unsupported actionable advice.

Incorrect Answers

A:

A fluent generation cannot substitute for missing authoritative transaction data. In the stated case, troubleshooting report: The model gives unverified treatment advice without escalation; “Propagate typed tool errors and use an explicit unavailable response” does not target that failure.

C:

A prompt-level request cannot override a transaction policy. In the stated case, troubleshooting report: The model gives unverified treatment advice without escalation; “Gate the payment tool with verified role and amount checks” does not target that failure.

D:

Privacy includes preventing cross-session memory disclosure. In the stated case, troubleshooting report: The model gives unverified treatment advice without escalation; “Partition state by user and clear or secure session context” does not target that failure.

Question 503

A project needs the following AI behavior: A transcription application must treat silence differently from a negative user answer. What should the designer implement?

  1. Use supported voice and SSML prosody rate controls
  2. Run grounded summarization after speech recognition and preserve source segments
  3. Distinguish no speech detected from recognized verbal denial
  4. Apply named entity recognition and preserve detected spans and categories

Explanation

Correct Answer

C:

An empty recognition result is not semantically equivalent to the word no. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

Voice selection and prosody can influence delivery without changing the text. That is not sufficient here: An empty recognition result is not semantically equivalent to the word no.

B:

Transcription obtains words; summarization extracts decisions and tasks. That is not sufficient here: An empty recognition result is not semantically equivalent to the word no.

D:

NER identifies named spans rather than summarizing the email. That is not sufficient here: An empty recognition result is not semantically equivalent to the word no.

Question 504

An AI operations team is testing two distinct capabilities. Failure (1): Only the local file path is passed as a text string. Failure (2): Overlapping detections are counted twice. Which TWO specific corrections should the team apply?

  1. Attach the image through the documented multimodal input field
  2. Apply a justified threshold with review for uncertain detections
  3. Parse outputs by their documented feature-specific schema
  4. Map the task to a whole-image classifier
  5. Apply confidence and nonmaximum suppression or instance deduplication

Explanation

Correct Answers

A:

Correct for failure (1). A textual filename cannot convey pixels to a model. The proposed correction directly addresses that reported defect.

E:

Correct for failure (2). Object counts require resolving distinct instances, not trusting one scene label. It independently corrects the second defect.

Incorrect Answers

B:

Confidence scores require calibration and contextual acceptance policy. It does not complete the two decisions required in this question.

C:

Tags describe concepts while OCR reports character content. It does not complete the two decisions required in this question.

D:

Image-level classification is distinct from object localization. It does not complete the two decisions required in this question.

Question 505

An acceptance tester needs to validate: Use a versioned evaluation set with stable scoring and test inputs. Which observed result proves the intended behavior?

  1. Old and new prompts are scored on the same labeled tasks
  2. The agent retrieves an authorized relevant passage and cites it
  3. A trace shows the task finished and its output is available
  4. An out-of-scope request is handled under the approved restrictions

Explanation

Correct Answer

A:

Anecdotal success on different prompts cannot establish an improvement. The expected acceptance evidence is: Old and new prompts are scored on the same labeled tasks. This observation directly tests the stated behavior.

Incorrect Answers

B:

Tool descriptions guide selection, while backend rules enforce allowed sources. This observation does not establish the required result: Old and new prompts are scored on the same labeled tasks.

C:

Submission is not the same as successful completion. This observation does not establish the required result: Old and new prompts are scored on the same labeled tasks.

D:

High-priority instructions should describe approved behavior and escalation rules. This observation does not establish the required result: Old and new prompts are scored on the same labeled tasks.

Question 506

The approved design action is: Use audio extraction followed by grounded action-item analysis. Which test result would BEST establish that it worked?

  1. Each invoice record contains only its own line items and total
  2. The extracted steps map to the correct clips and source times
  3. Each action item traces to spoken source intervals
  4. Observed actions link to supported source intervals while unobserved short events are not claimed as detected

Explanation

Correct Answer

C:

Transcription and summarization are separate results with different validation. The expected acceptance evidence is: Each action item traces to spoken source intervals. This observation directly tests the stated behavior.

Incorrect Answers

A:

Document boundaries prevent values from unrelated invoices being joined. This observation does not establish the required result: Each action item traces to spoken source intervals.

B:

Multimodal extraction draws on visual and spoken evidence over time. This observation does not establish the required result: Each action item traces to spoken source intervals.

D:

Content Understanding samples approximately one frame per second, so brief or single-frame actions can be missed despite valid timestamps. This observation does not establish the required result: Each action item traces to spoken source intervals.

Question 507

An AI operations team is testing two distinct capabilities. Failure (1): The system adds fabricated citations when search returns no documents. Failure (2): Raw identifiable frames are publicly attached to all summaries. Which TWO specific corrections should the team apply?

  1. Remove the false attribution and require documentary consent
  2. Require evidence-backed citation construction and abstention on no results
  3. Mask or exclude unauthorized identity details before publishing
  4. Add an accessible name and keyboard activation to the control
  5. Add auditable explanation records and an appeals workflow

Explanation

Correct Answers

B:

Correct for failure (1). Honesty requires acknowledging when a claim cannot be substantiated. The proposed correction directly addresses that reported defect.

C:

Correct for failure (2). Privacy applies to image content as well as text transcripts. It independently corrects the second defect.

Incorrect Answers

A:

Transparency prohibits presenting invented testimony as genuine. It does not complete the two decisions required in this question.

D:

Inclusive design permits nonvisual navigation of the same task path. The stated pair of failures instead requires “Require evidence-backed citation construction and abstention on no results” and “Mask or exclude unauthorized identity details before publishing”; this option cannot substitute for both.

E:

Accountability includes traceable responsibility and appropriate contestability. It does not complete the two decisions required in this question.

Question 508

An Azure AI Fundamentals solution has this requirement: A team needs to choose a deployment region for a model. Which implementation most directly addresses it?

  1. Check regional model availability and data handling requirements before deployment
  2. Record input/output token usage by model and prompt version
  3. Use the supported run or response status and verify its terminal state
  4. Run adversarial model evaluations and verify that configured content-safety guardrails enforce the stated high-severity block criteria

Explanation

Correct Answer

A:

Not every model or feature is available in every location. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

B:

Usage telemetry reveals context-growth and unexpected repetition. That is not sufficient here: Not every model or feature is available in every location.

C:

Submission is not the same as successful completion. That is not sufficient here: Not every model or feature is available in every location.

D:

Adversarial testing must measure actual configured policy enforcement rather than treating a prompt instruction as a guaranteed safety boundary. That is not sufficient here: Not every model or feature is available in every location.

Question 509

Troubleshooting report: The app retries mutating operations without an operation ID. Which correction is MOST directly targeted to this failure?

  1. Rotate the key and load credentials through a governed secret mechanism
  2. Attach idempotency keys and verify tool receipts after reconnecting
  3. Provision proper identity and resource role bindings for the target tenant
  4. Enforce a typed response schema and fail invalid output

Explanation

Correct Answer

B:

This correction directly addresses the reported failure. Client retries should distinguish read requests from side-effecting tool calls. The targeted change is: Attach idempotency keys and verify tool receipts after reconnecting.

Incorrect Answers

A:

Credentials should not appear in repositories, prompt logs or client binaries. In the stated case, troubleshooting report: The app retries mutating operations without an operation ID; “Rotate the key and load credentials through a governed secret mechanism” does not target that failure.

C:

Copying code does not grant access to another tenant's resources. In the stated case, troubleshooting report: The app retries mutating operations without an operation ID; “Provision proper identity and resource role bindings for the target tenant” does not target that failure.

D:

Model output constraints help downstream parsing but still require validation. In the stated case, troubleshooting report: The app retries mutating operations without an operation ID; “Enforce a typed response schema and fail invalid output” does not target that failure.

Question 510

An Azure-based AI team must satisfy this objective: A chatbot provides outdated pricing because its prompt includes last month's catalog. Which design choice is best?

  1. Define required output sections and link citations to actual retrieved passages
  2. Require explicit confirmation and backend authorization before tool execution
  3. Retrieve current approved pricing evidence instead of embedding volatile facts in instructions
  4. Pass relevant conversation context or a verified summarized state within token limits

Explanation

Correct Answer

C:

Static prompts are unsuitable as a changing source of commercial truth. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

Explicit response structure makes evidence easier to verify. That is not sufficient here: Static prompts are unsuitable as a changing source of commercial truth.

B:

High-impact side effects need permissions independent of conversational language. That is not sufficient here: Static prompts are unsuitable as a changing source of commercial truth.

D:

Stateful conversations require supplying or retrieving relevant prior context. That is not sufficient here: Static prompts are unsuitable as a changing source of commercial truth.

Microsoft Azure AI Fundamentals (AI-901) Practice Tests