AI0-001 Practice Test 26 — 15 Questions

Question 1

During a penetration test, a security engineer discovers that an AI-powered chatbot can be tricked into revealing sensitive customer data by using specially crafted prompts. What type of attack is this, and what is the best mitigation?

Choose an answer to begin — your selection is scored in the full session.

Accepted Answer

Prompt injection attack; implement input validation and context sanitization. This is a prompt injection attack, where an attacker crafts inputs that cause the AI model to override its original instructions or constraints, leading to unintended behavior such as revealing sensitive data. The best mitigation is input validation and context sanitization, which filters or neutralizes malicious prompt content before it reaches the model, preventing the injection from succeeding.

Answer

Model inversion attack; apply differential privacy during training

Answer

Data poisoning attack; implement strict access controls

Answer

Membership inference attack; add noise to model outputs

Question 2

A healthcare AI startup is developing a diagnostic tool that uses patient data to predict disease risk. To comply with HIPAA and minimize privacy risks while still training accurate models, which privacy-preserving technique should they prioritize?

Accepted Answer

Differential privacy. Differential privacy is the correct choice because it provides a formal mathematical guarantee that the output of a model does not reveal whether any individual's data was included in the training set. This is essential for HIPAA compliance as it prevents re-identification attacks even when an adversary has auxiliary information. Unlike anonymization or pseudonymization, differential privacy adds calibrated noise to the training process or query results, ensuring strong privacy protection while preserving model utility.

Answer

Anonymization

Answer

Pseudonymization

Answer

Data minimization

Question 3

A team is training a neural network for image classification. They observe that training loss decreases steadily but validation loss starts increasing after 20 epochs. What is the most likely issue?

Accepted Answer

Overfitting. The training loss decreasing while validation loss increases after 20 epochs is the classic signature of overfitting. The model is memorizing the training data (including noise) rather than learning generalizable patterns, causing it to perform poorly on unseen validation data.

Answer

Underfitting

Answer

Vanishing gradients

Answer

Data leakage

Question 4

A company implements an AI-based chatbot for customer service. After deployment, customers report that the chatbot sometimes uses offensive language. The development team reviews the training data and finds no explicit offensive content. What is the most likely explanation?

Accepted Answer

The model learned biased language patterns from the training corpus. The chatbot's offensive language likely stems from biased or toxic patterns present in the training corpus, even if no explicit offensive content was flagged. Large language models learn statistical associations from their training data, and if the corpus contains subtle biases, stereotypes, or indirect toxic language, the model can reproduce these patterns in its responses. This is a well-known issue in AI ethics and governance, where models inadvertently amplify societal biases embedded in the data.

Answer

There is a bug in the deployment pipeline

Answer

The model is overfitting to rare examples

Answer

The training data was poisoned by an attacker

Question 5

A security team is threat modeling an AI system that recommends financial products. They want to analyze threats unique to the ML pipeline using STRIDE. Which threat is LEAST applicable to the data collection and preprocessing stage?

Accepted Answer

Denial of Service (DoS). Denial of Service (DoS) is least applicable to the data collection and preprocessing stage because DoS threats typically target the availability of the model serving infrastructure (e.g., API endpoints, inference servers) rather than the static data ingestion pipeline. In the ML pipeline, DoS is more relevant during model deployment and inference, where an attacker could overwhelm the system with requests, not during the collection or preprocessing of training data.

Answer

Tampering with training data

Answer

Spoofing of data sources

Answer

Information disclosure via data leakage

Question 6

A healthcare organization uses an AI model to predict patient readmission risk. To comply with patient privacy regulations, they apply differential privacy during training. What is the primary trade-off of using differential privacy?

Accepted Answer

Reduced model accuracy for increased privacy. Differential privacy works by adding calibrated noise to the training process or model outputs, which directly reduces the model's accuracy in exchange for a quantifiable privacy guarantee (e.g., ε-differential privacy). This trade-off is fundamental: stronger privacy (lower ε) requires more noise, which degrades predictive performance. The healthcare organization must balance the need to protect patient data against the clinical utility of accurate readmission predictions.

Answer

Increased training time for reduced bias

Answer

Lower interpretability for higher fairness

Answer

Faster inference for lower memory usage

Question 7

A team uses Apache Kafka to stream real-time sensor data for ML inference. They need to process the stream, perform feature engineering, and store results in a data lake. Which tool is best suited for this streaming ML pipeline?

Accepted Answer

Apache Spark with Structured Streaming. Apache Spark with Structured Streaming is best suited because it provides a unified, scalable engine for both stream processing and batch processing, enabling real-time feature engineering on Kafka streams and direct writing to a data lake (e.g., Parquet format in Amazon S3). Its micro-batch or continuous processing model integrates natively with Kafka, allowing exactly-once semantics and low-latency transformations for ML inference pipelines.

Answer

Apache Airflow

Answer

TensorFlow Data Validation

Answer

SageMaker Processing jobs

Question 8

A machine learning team notices that their model's performance degrades when deployed to a new geographic region. The data distribution in the new region differs from the training data. Which concept best describes this issue?

Accepted Answer

Covariate shift. Covariate shift occurs when the distribution of the input features (covariates) changes between training and deployment, while the conditional relationship P(Y|X) remains the same. In this scenario, the model's performance degrades because the new geographic region has a different data distribution than the training data, which is the classic definition of covariate shift. This is a common issue in machine learning when models are deployed in environments not represented in the training set.

Answer

Data leakage

Answer

Underfitting

Answer

Overfitting

Question 9

A dataset for a binary classification problem has 95% of samples in class "0" and 5% in class "1". The data scientist trains a logistic regression model and achieves 95% accuracy. Which metric should the scientist primarily use to evaluate model performance?

Accepted Answer

Precision, recall, and F1-score.. In a highly imbalanced dataset (95% class 0, 5% class 1), accuracy is misleading because a model can achieve 95% accuracy by simply predicting the majority class for all samples. Precision, recall, and F1-score provide a more nuanced view of performance on the minority class, which is typically the class of interest in binary classification problems. The F1-score, in particular, balances precision and recall, making it the primary metric for evaluating model effectiveness on imbalanced data.

Answer

R-squared.

Answer

Accuracy.

Answer

Mean squared error.

Question 10

A machine learning team is splitting a dataset for a binary classification problem. They want to ensure robust evaluation and avoid data leakage. Which TWO practices should they follow? (Choose 2)

Accepted Answer

Split into training, validation, and test sets. Train/validation/test split is standard; cross-validation gives more robust estimates. Shuffling before split prevents ordering bias.

Answer

Normalise the entire dataset before splitting

Answer

Include validation data in the training set for more data

Answer

Use the same split for all experiments

Question 11

A machine learning team is evaluating a logistic regression model for a binary classification task. The dataset has 1,000 samples and 20 features. Which TWO metrics are most appropriate for evaluating model performance? (Choose 2)

Accepted Answer

F1 score. The F1 score is appropriate because it balances precision and recall, making it robust for binary classification when class distribution may be imbalanced. AUC-ROC measures the model's ability to distinguish between positive and negative classes across all classification thresholds, providing a threshold-independent evaluation of discriminative performance.

Answer

BLEU score

Answer

Perplexity

Answer

Accuracy

Question 12

An e-commerce company uses a machine learning model to recommend products to users. The model is retrained weekly and deployed to production. For the past three weeks, the model's click-through rate (CTR) has been stable except on Mondays, when it drops by 15%. Analysis reveals that the training data is extracted on Sundays and includes only weekday behavior. On Mondays, user behavior shifts due to weekend browsing patterns not captured in the training data. The team wants to maintain a weekly retraining cadence but fix the Monday performance drop. Which solution best addresses the Monday CTR drop without changing the retraining frequency?

Accepted Answer

Modify the data pipeline to include the full week (including the past weekend) in each retraining. It directly addresses the root cause: the training data excludes weekend behavior, causing the model to be blind to Monday patterns. By modifying the data pipeline to include the full week (including the past weekend) in each retraining, the model learns from weekend browsing patterns and can generalize to Monday user behavior without changing the weekly retraining cadence. This ensures the training distribution matches the inference distribution on Mondays, stabilizing CTR.

Answer

Deploy a separate model specifically for Monday predictions

Answer

Serve the previous week's model on Mondays to use older but stable patterns

Answer

Change to daily retraining to include weekend data more promptly

Question 13

A retail company wants to implement a recommendation system using collaborative filtering. The dataset contains user-item interactions (ratings) for 10,000 users and 5,000 products. The matrix is very sparse (99% missing values). The team plans to use matrix factorization to predict missing ratings. However, the training time is excessively long, and the model is not converging. The data engineer suggests using a smaller learning rate and more iterations. Which additional technique should the team apply to speed up training and improve convergence?

Accepted Answer

Add L2 regularization to the loss function. Adding L2 regularization to the loss function helps prevent overfitting and improves convergence in matrix factorization, especially with extremely sparse data (99% missing). Regularization penalizes large latent factor weights, which stabilizes the optimization process and allows the model to generalize better, reducing the risk of divergence during training.

Answer

Increase the minibatch size

Answer

Reduce the number of latent factors

Answer

Switch to the Adam optimizer

Question 14

Which TWO are common types of adversarial attacks on AI models?

Accepted Answer

Evasion attack. Evasion attacks (Option C) are a common type of adversarial attack where an attacker crafts malicious input data that is intentionally designed to cause a trained AI model to make incorrect predictions or classifications, often by adding imperceptible perturbations to legitimate inputs. This exploits the model's sensitivity to small changes in feature space, leading to misclassification without altering the model itself.

Answer

Hyperparameter tuning

Answer

Transfer learning

Answer

Backdoor attack

Question 15

Which TWO of the following are common techniques to improve the transparency and interpretability of an AI model?

Accepted Answer

Generate SHAP (SHapley Additive exPlanations) values. SHAP values are correct because they provide a unified measure of feature importance based on cooperative game theory, specifically Shapley values, which quantify the marginal contribution of each feature to a model's prediction. This makes the model's decision-making process transparent by showing how each input feature influences the output, which is a core technique for interpretability in AI governance.

Answer

Use differential privacy to add noise to training data

Answer

Implement a random forest algorithm

Answer

Use deep neural networks to increase model complexity