Azure AI Fundamentals Practice · Exam last updated: 19 Sept 2026
Total Questions
20
Question: 1
Certiva
A hospital chatbot must refuse to invent drug dosages. Which principle should the team prioritize?
Answer: B
Explanation:
Safety and human oversight beat a fluent but fabricated dosage. Keep a clinician in the loop.
Question: 2
Certiva
A retailer wants product-image captions generated at scale. Which service style fits that workload?
Answer: B
Explanation:
Vision captioning is an image-understanding job, not a speech-to-text or search-index job.
Question: 3
Certiva
A model scores loan applications. What is the most responsible next step before launch?
Answer: B
Explanation:
Measure disparate impact and document the decision policy. Shipping untested scoring is not responsible AI.
Question: 4
Certiva
Users dictate notes on a factory floor. Which capability matches the constraint?
Answer: B
Explanation:
Speech-to-text is the matching service when the input is audio and the output is text.
Question: 5
Certiva
A support bot must not reveal another customer’s ticket. Which control is essential?
Answer: B
Explanation:
Authorization and data isolation prevent cross-tenant leakage, even if the model is fluent.
Question: 6
Certiva
A team wants to classify support emails into billing vs. outage. What is the simplest fit?
Answer: B
Explanation:
Text classification maps labeled categories. It is not a video indexer or a DNS change.
Question: 7
Certiva
Which statement about training data is true for a workplace copilot?
Answer: B
Explanation:
You need rights to the data and a path to remove sensitive records. Scraping random web pages is not a license.
Question: 8
Certiva
An app translates product manuals into five languages. Which capability is the match?
Answer: B
Explanation:
Document or text translation is the dedicated capability; a firewall is not.
Question: 9
Certiva
A model might output copyrighted lyrics. What should the product do?
Answer: B
Explanation:
Block or cite restricted content and keep a human appeal path. Silent regurgitation is a risk.
Question: 10
Certiva
Which metric best tracks whether a classifier is guessing a rare class?
Answer: B
Explanation:
Precision/recall on the rare class beats overall accuracy when the set is imbalanced.
Question: 11
Certiva
A kiosk reads handwritten forms. Which Azure-style capability applies?
Answer: B
Explanation:
OCR / document intelligence extracts print and handwriting. A load balancer does not.
Question: 12
Certiva
Stakeholders ask why a model denied a claim. What should you provide?
Answer: B
Explanation:
An explanation of features and a human review path. “The neural net said so” is not enough.
Question: 13
Certiva
A chatbot must stay on-policy for a regulated bank. Which pattern helps?
Answer: B
Explanation:
Ground answers in approved retrieval sources and refuse when sources are missing.
Question: 14
Certiva
Which data class should never be pasted into an unmanaged prompt?
Answer: B
Explanation:
Secrets, government IDs, and health records need a governed path, not a public playground.
Question: 15
Certiva
A team detects drift after a catalog change. What is the first operational step?
Answer: B
Explanation:
Monitor, compare against a baseline, and re-evaluate. Ignoring drift lets quality decay.
Question: 16
Certiva
You need to extract key-value pairs from invoices. Which capability fits?
Answer: B
Explanation:
Document intelligence / form extraction is built for invoices. A CDN is not.
Question: 17
Certiva
Which human-in-the-loop design is appropriate for refund decisions over a threshold?
Answer: B
Explanation:
High-impact money movement should pause for a reviewer. Full automation of large refunds is risky.
Question: 18
Certiva
A voice agent must detect a wake phrase in noisy audio. What is the matching skill?
Answer: B
Explanation:
Keyword spotting / speech recognition with noise handling, not a packet sniffer.
Question: 19
Certiva
What is a valid reason to decline generating an image of a real private individual?
Answer: B
Explanation:
Likeness and consent. Public clip-art of objects is a different case.
Question: 20
Certiva
A search experience should rank internal wiki pages. Which pattern applies?
Answer: B
Explanation:
Semantic or hybrid search over your corpus, with access control, is the fit.