- A
The difference in model accuracy between a private deployment and a public API
Why wrong: Performance differences are deployment metrics — differential privacy is a mathematical privacy guarantee for training on sensitive data.
- B
A mathematical guarantee that model training reveals negligible information about any individual's data
Differential privacy adds noise during training — providing formal guarantees that models don't memorise or expose individual records.
- C
Encrypting model weights so they remain private from users accessing the model API
Why wrong: Model weight encryption is IP protection — differential privacy protects training data individuals, not the model weights themselves.
- D
Using different models for different privacy tiers of customers
Why wrong: Tiered model access is a product design decision — differential privacy is a specific mathematical technique for privacy-preserving training.
Differential Privacy: Protecting Individual Data in Training
This AI-900 practice question tests your understanding of describe fundamental principles of machine learning on azure. Read the scenario carefully and evaluate each option against the stated constraints before committing to an answer. After answering, compare your reasoning against the explanation and wrong-answer breakdown below. Once you have made your selection, read the full explanation to reinforce the concept and understand why each distractor is designed to mislead on exam day.
What is 'differential privacy' and how is it relevant to AI model training?
Quick Answer
The answer is a mathematical guarantee that model training reveals negligible information about any individual’s data. This is correct because differential privacy works by adding calibrated noise to the training process or query results, ensuring that the output does not change significantly whether a specific person’s data is included or excluded—a formal privacy guarantee quantified by the epsilon parameter. On the AI-900 exam, this concept tests your understanding of how Azure enables responsible AI by protecting individual privacy during model training, often appearing in questions about data governance or compliance with regulations like GDPR and HIPAA. A common trap is confusing differential privacy with data anonymization; remember that anonymization removes identifiers, while differential privacy adds noise to the training itself. Memory tip: think “epsilon noise” to recall that differential privacy injects controlled randomness to shield individual data points.
Answer choices
Why each option matters
Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.
Correct answer & explanation
A mathematical guarantee that model training reveals negligible information about any individual's data
Differential privacy is a mathematical framework that ensures the output of a model training process does not reveal whether any specific individual's data was included in the training dataset. It achieves this by adding calibrated noise to the training process or query results, providing a formal privacy guarantee quantified by the epsilon parameter. This is directly relevant to AI model training because it allows organizations to train models on sensitive data while protecting individual privacy, which is a core requirement for compliance with regulations like GDPR and HIPAA.
Key principle: Answer the scenario, not the keyword: identify the specific constraint before choosing the most familiar-sounding option.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
The difference in model accuracy between a private deployment and a public API
Why it's wrong here
Performance differences are deployment metrics — differential privacy is a mathematical privacy guarantee for training on sensitive data.
- ✓
A mathematical guarantee that model training reveals negligible information about any individual's data
Why this is correct
Differential privacy adds noise during training — providing formal guarantees that models don't memorise or expose individual records.
Related concept
Read the scenario before looking for a memorised answer.
- ✗
Encrypting model weights so they remain private from users accessing the model API
Why it's wrong here
Model weight encryption is IP protection — differential privacy protects training data individuals, not the model weights themselves.
- ✗
Using different models for different privacy tiers of customers
Why it's wrong here
Tiered model access is a product design decision — differential privacy is a specific mathematical technique for privacy-preserving training.
Common exam traps
Common exam trap: answer the scenario, not the keyword
The trap here is that candidates confuse data privacy techniques (like encryption or access control) with the formal mathematical guarantee of differential privacy, which specifically addresses information leakage from the model's outputs rather than protecting the data at rest or in transit.
Detailed technical explanation
How to think about this question
Under the hood, differential privacy works by adding random noise (e.g., Laplace or Gaussian noise) to the gradients during stochastic gradient descent or to the final model parameters, calibrated to the sensitivity of the query and the desired epsilon budget. A subtle behavior is that the privacy guarantee holds even if an adversary has access to all other data points except the target individual's, making it robust against differencing attacks. In a real-world scenario, Azure Machine Learning integrates differential privacy via the SmartNoise toolkit, allowing data scientists to train models on medical records while provably limiting the risk of re-identification.
KKey Concepts to Remember
- Read the scenario before looking for a memorised answer.
- Find the constraint that changes the correct option.
- Eliminate answers that are true in general but not in this case.
TExam Day Tips
- Watch for words such as best, first, most likely and least administrative effort.
- Review why wrong options are wrong, not only why the correct option is correct.
Key takeaway
Answer the scenario, not the keyword: identify the specific constraint before choosing the most familiar-sounding option.
Real-world example
How this comes up in practice
A media company stores terabytes of video archives that are accessed once a year for audit purposes. Moving these objects to a cold storage tier (Azure Archive, S3 Glacier, or Google Nearline) costs a fraction of hot storage. Questions like this test whether you understand storage tiers, access frequency tradeoffs, and retrieval latency requirements.
What to study next
Got this wrong? Here's your next step.
Identify which exam domain this question belongs to, review the core concept, then practise similar questions from the same domain.
- →
Describe fundamental principles of machine learning on Azure — study guide chapter
Learn the concepts, then practise the questions
- →
Describe fundamental principles of machine learning on Azure practice questions
Targeted practice on this topic area only
- →
All AI-900 questions
1,020 questions across all exam domains
- →
Microsoft Azure AI Fundamentals AI-900 study guide
Full concept coverage aligned to exam objectives
- →
AI-900 practice test guide
How to use practice tests most effectively before exam day
Related practice questions
Related AI-900 practice-question pages
Use these pages to review the topic behind this question. This is how one missed question becomes focused revision.
Describe Artificial Intelligence workloads and considerations practice questions
Practise AI-900 questions linked to Describe Artificial Intelligence workloads and considerations.
Describe fundamental principles of machine learning on Azure practice questions
Practise AI-900 questions linked to Describe fundamental principles of machine learning on Azure.
Describe features of computer vision workloads on Azure practice questions
Practise AI-900 questions linked to Describe features of computer vision workloads on Azure.
Describe features of Natural Language Processing workloads on Azure practice questions
Practise AI-900 questions linked to Describe features of Natural Language Processing workloads on Azure.
Describe features of generative AI workloads on Azure practice questions
Practise AI-900 questions linked to Describe features of generative AI workloads on Azure.
AI-900 fundamentals practice questions
Practise AI-900 questions linked to AI-900 fundamentals.
AI-900 scenario practice questions
Practise AI-900 questions linked to AI-900 scenario.
AI-900 troubleshooting practice questions
Practise AI-900 questions linked to AI-900 troubleshooting.
Practice this exam
Start a free AI-900 practice session
Short sessions build daily habit. Longer sessions build exam-day stamina. Try a timed session to simulate real conditions.
FAQ
Questions learners often ask
What does this AI-900 question test?
Describe fundamental principles of machine learning on Azure — This question tests Describe fundamental principles of machine learning on Azure — Read the scenario before looking for a memorised answer..
What is the correct answer to this question?
The correct answer is: A mathematical guarantee that model training reveals negligible information about any individual's data — Differential privacy is a mathematical framework that ensures the output of a model training process does not reveal whether any specific individual's data was included in the training dataset. It achieves this by adding calibrated noise to the training process or query results, providing a formal privacy guarantee quantified by the epsilon parameter. This is directly relevant to AI model training because it allows organizations to train models on sensitive data while protecting individual privacy, which is a core requirement for compliance with regulations like GDPR and HIPAA.
What should I do if I get this AI-900 question wrong?
Identify which exam domain this question belongs to, review the core concept, then practise similar questions from the same domain.
What is the key concept behind this question?
Read the scenario before looking for a memorised answer.
About these practice questions
Courseiva creates original exam-style practice questions with explanations and wrong-answer analysis. It does not publish real exam questions, exam dumps, or protected exam content. Learn why practice questions differ from exam dumps →
Keep practising
More AI-900 practice questions
- A company deploys an AI system to screen job applications. The system is a complex neural network that learns patterns f…
- What is 'model versioning' and why is it essential in MLOps?
- What is 'AI transparency' in Microsoft's Responsible AI principles?
- A company uses Azure OpenAI Service to generate marketing copy. They notice that sometimes the generated text contains r…
- A data scientist is training a regression model to predict house prices using features like square footage, number of be…
- A company uses Azure OpenAI Service to generate marketing copy. They want to ensure that the generated text does not con…
Last reviewed: Jun 11, 2026
This AI-900 practice question is part of Courseiva's free Microsoft certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the AI-900 exam.
Question Discussion
Share a tip, memory trick, or ask about the reasoning behind this question. Do not post real exam questions, leaked content, braindumps, or copyrighted exam material. Comments are moderated and may be removed without notice.
Sign in to join the discussion.