Home/microsoft/Free Microsoft AI-900 Azure AI Fundamentals Actual Exam Questions

Free Microsoft AI-900 Azure AI Fundamentals Actual Exam Questions

The questions for this exam were last updated on January 9, 2026

Dumps Box (DumpsBox) offers up-to-date practice exam questions for AI-900 certification exam which are developed and validated by Microsoft subject domain experts certified in Microsoft AI-900 Azure AI Fundamentals . These practice questions are update regularly as we keep an eye on any recent changes in AI-900 syllabus, and when there is update our team quickly adjusts the questions. This commitment to providing the best quality exam prep material to certification aspirants is what makes DumpsBox.com the best certification exam prep website. On top of that, our strong, yet strictly moderated, community based feedback keeps the content clean and current. Each question has helpful community discussion that provides it extra perspective and introduces helpful resources for better exam preparation. This also saves students from other outdated practice questions or illicit exam dumps that can have adverse affects on career. Browse through our Microsoft AI-900 Azure AI Fundamentals exam questions and pass your exam on first try.

Question No. 1

HOTSPOT Select the answer that correctly completes the sentence. AI-900 practice exam questions

Top comments
RU
Ryan U.
2026-02-19

C, since scalability means handling growth, not just cost savings.

0
RU
Ryan U.
2026-02-17

C, because it’s about managing bigger workloads, not just reducing expenses.

0
Question No. 2
Your company wants to build a recycling machine for bottles. The recycling machine must
automatically identify bottles of the correct shape and reject all other items.
Which type of AI workload should the company use?
Select one option, then reveal solution.
Top comments
AH
Ash H.
2026-01-19

Isn’t this just classifying images? So C?

0
Question No. 3Drag & Drop

DRAG DROP Match the tool to the Azure Machine Learning task. To answer, drag the appropriate tool from the column on the left to its tasks on the right. Each tool may be used once, more than once, or not at all NOTE: Each correct match is worth one point. AI-900 practice exam questions

Options
AAutomated machine learning (automated ML)
BThe Azure portal
CMachine Learning designer
Drag an item to a target. Click × to remove.
Answer Area
Create a Machine Learning workspace
Drop item here
Use a drag-and-drop interface used to train and deploy models
Drop item here
Use a wizard to select configurations for a machine learning run
Drop item here
Top comments
CE
Carlos E.
2026-02-22

I matched Automated ML with model creation since it automates the process. Designer fits with pipeline building. Notebooks are for custom experiments and coding tasks. This fits their core purposes clearly.

0
DV
David V.
2026-02-11

I also went with automated ML for model creation since it’s designed to streamline that. For experimentation, Azure Notebooks makes sense because it’s more hands-on with code. The designer is more about building models visually without coding, so it fits both creation and deployment prep. Some options don’t seem to match anything here, like if there’s a tool focused purely on monitoring or management. I figured the key is focusing on what each tool’s primary strength is rather than trying to fit them all in multiple tasks.

0
Question No. 4

HOTSPOT To complete the sentence, select the appropriate option in the answer area. AI-900 practice exam questions

Top comments
YY
Yasir Y.
2026-02-13

I’m dropping option C here because it doesn’t mention the key update the question focuses on. Only B talks about that specific feature, so it makes the most sense.

0
SE
Sarah E.
2026-01-28

Option B sounds right since it highlights the newest feature clearly.

0
Question No. 5
Which parameter should you configure to produce a more diverse range of tokens in the responses
from a chat solution that uses the Azure OpenAI GPT-3.5 model?
Select one option, then reveal solution.
Top comments
KN
Kevin N.
2026-02-22

C imo, presence penalty directly affects repetition, making responses less predictable.

0
KN
Kevin N.
2026-02-20

Option C seems right because presence penalty discourages repeating the same words, which naturally leads to more varied responses. The other options don’t really influence token diversity directly.

0
Question No. 6

HOTSPOT For each of the following statements, select Yes if the statement is true. Otherwise, select No. NOTE: Each correct selection is worth one point. AI-900 practice exam questions

Top comments
SE
Sarah E.
2026-02-19

B and D are about actual benefits, so they seem right.

0
BF
Brian F.
2026-02-10

I’m seeing B and D as the only realistic yes answers since they talk about AI improving efficiency and insights. A and C feel more about fears than actual benefits or truths.

0
Question No. 7Drag & Drop

DRAG DROP Match the Microsoft guiding principles for responsible AI to the appropriate descriptions. To answer, drag the appropriate principle from the column on the left to its description on the right. Each principle may be used once, more than once, or not at all. NOTE: Each correct selection is worth one point. AI-900 practice exam questions

Options
AAccountability
BFairness
CInclusiveness
DPrivacy and security
EReliability and safety
Drag an item to a target. Click × to remove.
Answer Area
Ensure that Al systems operate as they were originally designed, respond to unanticipated conditions, and resist harmful manipulation.
Drop item here
Implementing processes to ensure that decisions made by Al systems can be overridden by humans.
Drop item here
Provide consumers with information and controls over the collection, use, and storage of their data.
Drop item here
Top comments
SO
Sam O.
2026-02-20

I’d match fairness with promoting equality, since it’s about bias and inclusion.

0
SO
Sam O.
2026-02-09

I’d pair privacy with protecting user data since it’s all about confidentiality. Then, reliability and safety seem to go together with making sure the AI system works consistently and avoids harm. For accountability, I agree it’s about audit trails and fixing problems. Transparency should be about explaining how decisions are made clearly. Fairness fits naturally with avoiding any kind of bias or discrimination. Just matching the principle’s main focus helps eliminate confusion quickly.

0
Question No. 8
Which two tools can you use to call the Azure OpenAI service? Each correct answer presents a
complete solution.
NOTE: Each correct answer is worth one point.
Select all that apply, then reveal solution.
Top comments
SA
Sohail A.
2026-02-22

C imo, since SDKs provide full programmatic access and handle authentication better than CLI. D also works for the same reason, letting JS devs call the service smoothly without dealing with REST details.

0
SA
Sohail A.
2026-02-18

Probably A and B. CLI is straightforward for quick commands, and REST API is the base method that everything else builds on, so both definitely count as complete solutions.

0
Question No. 9
For a machine learning progress, how should you split data for training and evaluation?
Select one option, then reveal solution.
Top comments
JT
John T.
2026-02-20

B/D? Splitting columns (D) would mix up features and labels, so that’s a no. B makes sense because it keeps the samples intact, which is crucial for training and evaluation.

0
AD
Amir D.
2026-02-17

Probably B, since splitting rows keeps the feature-label pairs intact for both training and evaluation. Splitting columns (D) would separate features and mess up the data.

0
Question No. 10
You need to develop a chatbot for a website. The chatbot must answer users questions based on the
information m the following documents
• A product troubleshooting guide m a Microsoft Word document
• A frequently asked questions (FAQ) list on a webpage
Which service should you use to process the documents?
Select one option, then reveal solution.
Top comments
OV
Omar V.
2026-02-18

Maybe D, since QnA Maker is designed to pull Q&A from different documents and webpages, which fits the need to combine both the Word guide and FAQ list effectively.

0
WU
Will U.
2026-02-12

It’s D because QnA Maker specializes in extracting Q&A pairs from various sources, making it easier to integrate both the Word guide and the FAQ page into one searchable chatbot.

0
Question No. 11

HOTSPOT For each of the following statements, select Yes if the statement is true. Otherwise, select No. NOTE; Each correct selection is worth one point. AI-900 practice exam questions

Top comments
RT
Rizwan T.
2026-02-13

I agree with ruling out B and D here. B mentions features that aren’t really part of Azure Machine Learning Studio’s core functionality, and D seems to be describing something more like a cognitive service rather than ML Studio. A and C both directly relate to what ML Studio offers—building and deploying models and managing experiments. So, focusing just on Azure Machine Learning Studio, A and C should be the picks.

0
AI
Adeel I.
2026-01-27

B can be ruled out since it talks about capabilities not available in Azure Machine Learning Studio. D also feels off because it's about a different service entirely. So yeah, A and C make the most sense here.

0
Question No. 12
You use Azure Machine Learning designer to publish an inference pipeline.
Which two parameters should you use to consume the pipeline? Each correct answer presents part
of the solution.
NOTE: Each correct selection is worth one point.
Select all that apply, then reveal solution.
Top comments
AP
Arjun P.
2026-02-17

It’s definitely D and C for me. The REST endpoint is the URL you call to send data for inference, and the authentication key is necessary to access that endpoint securely. The model name doesn’t really matter once the pipeline is published, and the training endpoint is only relevant during training, not for consuming the pipeline.

0
AP
Arjun P.
2026-02-13

Guessing C and D make the most sense here since the REST endpoint is where you send requests, and the auth key secures access. The model name and training endpoint seem unrelated to calling the published pipeline.

0
Question No. 13
A medical research project uses a large anonymized dataset of brain scan images that are categorized
into predefined brain haemorrhage types.
You need to use machine learning to support early detection of the different brain haemorrhage
types in the images before the images are reviewed by a person.
This is an example of which type of machine learning?
Select one option, then reveal solution.
Top comments
TG
Tom G.
2026-02-22

It’s C because the categories are known, so it’s definitely supervised classification.

0
SB
Sam B.
2026-02-20

A/B? I ruled out regression (B) since this isn’t about predicting a number or continuous value. Clustering (A) is unsupervised, but here we have predefined categories, so that doesn’t fit either.

0
Question No. 14
You need to generate images based on user prompts. Which Azure OpenAI model should you use?
Select one option, then reveal solution.
Top comments
FN
Farhan N.
2026-02-18

Maybe D is easy to rule out since Whisper is for speech-to-text, not images. GPT-3 and GPT-4 are mainly language models, so they don’t fit image generation either. That leaves B, DALL-E, which is designed for creating images from prompts, so it makes the most sense here.

0
FN
Farhan N.
2026-01-29

B imo, since GPT-4 and GPT-3 are mainly for text tasks, and Whisper is for audio transcription. DALL-E is specifically designed for creating images from text.

0
Question No. 15

HOTSPOT You have the following apps: • App1: Uses a set of images and photos to extract brand names • App2: Enables touchless access control for buildings Which Azure Al Vision service does each app use? To answer, select the appropriate options in the answer area. NOTE: Each correct selection is worth one point. AI-900 practice exam questions

Top comments
FK
Farhan K.
2026-02-20

For App1, I think it’s less about just reading text and more about identifying specific brands, so Custom Vision makes sense since it’s trained for custom image classification. The Read API would be more general OCR, but here the goal is brand recognition. As for App2, touchless access control likely involves identifying people or verifying identities, so Face API fits better than just general object detection. So I’d pair App1 with Custom Vision and App2 with Face API based on their core functions.

0
BF
Brian F.
2026-02-10

App1 is probably using the Read API because it's extracting text (brand names) specifically from images. Even though Custom Vision handles image classification, extracting text is more about OCR capabilities, which the Read API does. For App2, touchless access suggests facial recognition or identity verification, so Face API fits better than Video Analyzer since it focuses on people recognition rather than video content analysis. So I’d pick Read API for App1 and Face API for App2.

0