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 40

Question 390

An Azure AI Fundamentals solution has this requirement: A safety filter allows an answer after retrieving a malicious document. Which implementation most directly addresses it?

  1. Evaluate outcomes with counterfactual name swaps and remove unjustified proxy effects
  2. Test representative document layouts and accessibility-oriented inputs
  3. Evaluate safety again after tool and retrieval results are incorporated
  4. Use grounded medical content, high-risk detection and appropriate expert escalation

Explanation

Correct Answer

C:

Threats can enter later stages even when the original user request was harmless. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

Fairness checks compare materially equivalent cases rather than raw group averages alone. That is not sufficient here: Threats can enter later stages even when the original user request was harmless.

B:

Inclusion requires robust input handling beyond a single document template. That is not sufficient here: Threats can enter later stages even when the original user request was harmless.

D:

Safety mechanisms should reduce unsupported high-impact recommendations. That is not sufficient here: Threats can enter later stages even when the original user request was harmless.

Question 391

An Azure AI solution uses the following approved control: Request an image caption or supported multimodal description. What observable result should pass QA?

  1. A labeled test shows the chosen threshold achieves target precision/recall
  2. The thumbnail preserves subject proportions and important content
  3. The output is a new raster image matching the approved icon brief
  4. A labeled photograph receives an accurate high-level summary

Explanation

Correct Answer

D:

Captioning produces a sentence-level description rather than a coordinate list. The expected acceptance evidence is: A labeled photograph receives an accurate high-level summary. This observation directly tests the stated behavior.

Incorrect Answers

A:

Confidence scores require calibration and contextual acceptance policy. This observation does not establish the required result: A labeled photograph receives an accurate high-level summary.

B:

Smart cropping differs from simply stretching pixels. This observation does not establish the required result: A labeled photograph receives an accurate high-level summary.

C:

Image generation creates new pixels rather than classifying an existing photo. This observation does not establish the required result: A labeled photograph receives an accurate high-level summary.

Question 392

An acceptance tester needs to validate: Recompute arithmetic and compare totals with reported invoice fields. Which observed result proves the intended behavior?

  1. Validated fields above tested thresholds proceed while uncertain or inconsistent fields enter review
  2. The deployed analyzer uses a documented production-capable version
  3. Extracted line items reconcile to visible row amounts
  4. Total mismatches trigger a review rather than automatic posting

Explanation

Correct Answer

D:

Independent business rules catch plausible extraction errors. The expected acceptance evidence is: Total mismatches trigger a review rather than automatic posting. This observation directly tests the stated behavior.

Incorrect Answers

A:

Content Understanding confidence applies to supported document analyzers; confidence estimates alone do not prove field correctness. This observation does not establish the required result: Total mismatches trigger a review rather than automatic posting.

B:

Preview-only features may have different availability and contracts. This observation does not establish the required result: Total mismatches trigger a review rather than automatic posting.

C:

Cell relationships are necessary to check item totals. This observation does not establish the required result: Total mismatches trigger a review rather than automatic posting.

Question 393

Troubleshooting report: Raw prompts and full identifiers appear in debug output. Which correction is MOST directly targeted to this failure?

  1. Add adversarial and exception-focused cases before promotion
  2. Remove protected payloads from logging and enforce redaction tests
  3. Apply real sign-in and role checks before invoking protected tools
  4. Add subgroup and matched-pair fairness tests before approval

Explanation

Correct Answer

B:

This correction directly addresses the reported failure. Troubleshooting does not justify reproducing sensitive application records. The targeted change is: Remove protected payloads from logging and enforce redaction tests.

Incorrect Answers

A:

A high overall score can conceal severe failure on unusual inputs. In the stated case, troubleshooting report: Raw prompts and full identifiers appear in debug output; “Add adversarial and exception-focused cases before promotion” does not target that failure.

C:

Natural-language claims cannot elevate authentication authority. In the stated case, troubleshooting report: Raw prompts and full identifiers appear in debug output; “Apply real sign-in and role checks before invoking protected tools” does not target that failure.

D:

Fairness checks compare materially equivalent cases rather than raw group averages alone. In the stated case, troubleshooting report: Raw prompts and full identifiers appear in debug output; “Add subgroup and matched-pair fairness tests before approval” does not target that failure.

Question 394

An engineering review documents this requirement: A marketing analyst wants to find customer sentiment in product reviews. Select the most appropriate solution.

  1. Use anomaly detection against expected behavioral patterns
  2. Use a visual classification or detection model appropriate to whether location matters
  3. Use sentiment analysis to estimate positive, negative, or neutral tone
  4. Use an agent workflow with a permitted ticket tool and an authorization gate

Explanation

Correct Answer

C:

Sentiment analysis infers expressed attitude rather than extracting names. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

Anomaly methods identify deviations rather than assigning preset incident labels. That is not sufficient here: Sentiment analysis infers expressed attitude rather than extracting names.

B:

The task chooses a label, with localization needed only when multiple signs appear. That is not sufficient here: Sentiment analysis infers expressed attitude rather than extracting names.

D:

Agentic workloads combine reasoning and bounded external actions. That is not sufficient here: Sentiment analysis infers expressed attitude rather than extracting names.

Question 395

An acceptance tester needs to validate: Recognize speech, translate text, then synthesize French output. Which observed result proves the intended behavior?

  1. A spoken inquiry yields the correct audible answer after processing
  2. The listener hears equivalent French instructions
  3. The spoken answer slows while preserving its content
  4. The result contains central complaint phrases without invented terms

Explanation

Correct Answer

B:

Cross-language audio output requires both language transfer and synthesis. The expected acceptance evidence is: The listener hears equivalent French instructions. This observation directly tests the stated behavior.

Incorrect Answers

A:

Input transcription and output synthesis are distinct steps. This observation does not establish the required result: The listener hears equivalent French instructions.

C:

Voice selection and prosody can influence delivery without changing the text. This observation does not establish the required result: The listener hears equivalent French instructions.

D:

Key phrases highlight important concepts expressed in the original text. This observation does not establish the required result: The listener hears equivalent French instructions.

Question 396

The approved design action is: Measure expected tokens and requests per minute, then choose an approved throughput tier or capacity adjustment before launch. Which test result would BEST establish that it worked?

  1. Repeated structured cases follow the expected schema more consistently
  2. A peak-load forecast matches the configured token/request limits or identifies an approved capacity increase
  3. The model/deployment capability check and a spoken-audio test confirm that this endpoint supports transcription
  4. Unseen examples have measured results independent of training

Explanation

Correct Answer

B:

Quota sizing is a deployment-capacity decision; retrying a predictable sustained excess cannot add model throughput. The expected acceptance evidence is: A peak-load forecast matches the configured token/request limits or identifies an approved capacity increase. This observation directly tests the stated behavior.

Incorrect Answers

A:

Lower randomness often improves consistency but does not guarantee correctness. This observation does not establish the required result: A peak-load forecast matches the configured token/request limits or identifies an approved capacity increase.

C:

Model capabilities must include the required audio modality before the client can obtain speech-to-text results. This observation does not establish the required result: A peak-load forecast matches the configured token/request limits or identifies an approved capacity increase.

D:

A model can memorize training examples without generalizing. This observation does not establish the required result: A peak-load forecast matches the configured token/request limits or identifies an approved capacity increase.

Question 397

An acceptance tester needs to validate: Use image classification for the requested whole-image class. Which observed result proves the intended behavior?

  1. The output is a validated outdoor probability or category
  2. Restricted users cannot retrieve the original image
  3. The same object is recognized consistently across supported rotations
  4. A correctly converted test image processes successfully

Explanation

Correct Answer

A:

Image-level classification is distinct from object localization. The expected acceptance evidence is: The output is a validated outdoor probability or category. This observation directly tests the stated behavior.

Incorrect Answers

B:

Images can carry personally identifying information beyond text fields. This observation does not establish the required result: The output is a validated outdoor probability or category.

C:

Rotated inputs can degrade detection or coordinate interpretation. This observation does not establish the required result: The output is a validated outdoor probability or category.

D:

Vision endpoints impose supported input formats and limits. This observation does not establish the required result: The output is a validated outdoor probability or category.

Question 398

The approved design action is: Use text summarization with clear length and fidelity constraints. Which test result would BEST establish that it worked?

  1. A summary preserves issue, action and resolution from the ticket
  2. A spoken inquiry yields the correct audible answer after processing
  3. The sanitized SSML parses and renders the intended characters
  4. The transcript matches a labeled spoken test phrase

Explanation

Correct Answer

A:

Summarization compresses written information rather than changing its facts. The expected acceptance evidence is: A summary preserves issue, action and resolution from the ticket. This observation directly tests the stated behavior.

Incorrect Answers

B:

Input transcription and output synthesis are distinct steps. This observation does not establish the required result: A summary preserves issue, action and resolution from the ticket.

C:

SSML is XML and malformed markup breaks the synthesis request. This observation does not establish the required result: A summary preserves issue, action and resolution from the ticket.

D:

Speech recognition converts waveform features to text tokens. This observation does not establish the required result: A summary preserves issue, action and resolution from the ticket.

Question 399

An Azure-based AI team must satisfy this objective: A meeting recording needs speaker-attributed action items. Which design choice is best?

  1. Combine audio transcription, speaker segmentation and grounded summarization
  2. Use an agent with scoped retrieval and separately authorized mutation tools
  3. Use document information extraction with a typed field schema
  4. Use OCR tuned to the source quality and validate extracted text

Explanation

Correct Answer

A:

Attribution distinguishes who committed to each action. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

B:

Agentic workflows combine information access with controlled actions. That is not sufficient here: Attribution distinguishes who committed to each action.

C:

Document extraction maps unstructured layouts into structured values. That is not sufficient here: Attribution distinguishes who committed to each action.

D:

Text recognition reads characters rather than merely describing visual objects. That is not sufficient here: Attribution distinguishes who committed to each action.

Question 400

An Azure AI Fundamentals solution has this requirement: A PDF contains a three-column price table and a computed total. Which implementation most directly addresses it?

  1. Use timestamped multi-segment video analysis for visible actions, with separate higher-frame-rate review when brief security events matter
  2. Preserve page identifiers and validate page sequence during extraction
  3. Preserve table row-column structure and validate numeric totals
  4. Check feature, model and regional availability before creating an analyzer

Explanation

Correct Answer

C:

Cell relationships are necessary to check item totals. This implementation addresses the precise goal rather than an adjacent AI task.

Incorrect Answers

A:

Content Understanding samples approximately one frame per second, so brief or single-frame actions can be missed despite valid timestamps. That is not sufficient here: Cell relationships are necessary to check item totals.

B:

Page order and references are part of document provenance. That is not sufficient here: Cell relationships are necessary to check item totals.

D:

Some modalities and features depend on resource and API version. That is not sufficient here: Cell relationships are necessary to check item totals.

Microsoft Azure AI Fundamentals (AI-901) Practice Tests