AI0-001 Practice Test 26 — 15 Questions

Question 1

A company deploys an AI chatbot that generates product descriptions. The company wants to be transparent about AI-generated content. Which practice should they follow?

Accepted Answer

Clearly label AI-generated content as such. Transparency about AI-generated content is a core principle of AI governance and ethics. Labeling AI-generated outputs as such allows users to make informed decisions about the content they consume, aligning with responsible AI practices.

Answer

Publish a model card, but not label individual outputs

Answer

Add an invisible watermark but do not inform users

Answer

Do not disclose that content is AI-generated to avoid user confusion

Question 2

Which TWO actions should be taken to ensure an AI model complies with GDPR requirements when processing personal data?

Accepted Answer

Limit data collection to only what is necessary for the model. Option A is correct because GDPR's data minimization principle (Article 5(1)(c)) requires that personal data be adequate, relevant, and limited to what is necessary for the purposes for which it is processed, so restricting collection to only what the model needs directly supports compliance. Option E is correct because GDPR grants data subjects the right to erasure (Article 17, the 'right to be forgotten'), so implementing user data deletion upon request is a mandatory capability for any system processing personal data. Option B is not required by GDPR, which focuses on transparency about processing purposes and logic rather than mandating a full explanation of every model prediction (that concern relates more to interpretability and AI-specific regulation). Option C is wrong because GDPR's storage limitation principle requires data to be kept no longer than necessary, so a 10-year minimum retention period would violate the regulation. Option D is not strictly required because GDPR permits processing of personal data with a lawful basis; anonymization is one way to reduce risk but is not a universal prerequisite, and true anonymization is often impractical for model training.

Answer

Provide a full explanation of model predictions

Answer

Store all user data for a minimum of 10 years

Answer

Anonymize all personal data before use

Question 3

An AI system must extract text from scanned invoices and output structured fields (invoice number, date, total amount). Which type of AI application is this?

Accepted Answer

Document intelligence. Document intelligence (D) is the correct answer because it specifically refers to AI systems that extract, classify, and structure data from documents like invoices, receipts, and forms. This application uses optical character recognition (OCR) combined with natural language processing (NLP) to identify and output structured fields such as invoice number, date, and total amount, which is exactly what the question describes.

Answer

Chatbot/virtual assistant

Answer

Code generation

Answer

Image classification/object detection

Question 4

An AI practitioner needs to measure the performance of a binary classification model for disease detection, where the cost of false negatives is very high. Which metric should be prioritized?

Accepted Answer

Recall. Recall (true positive rate) minimises false negatives, which is critical when missing a positive case is dangerous.

Answer

Precision

Answer

F1-score

Answer

Accuracy

Question 5

A company wants to train a language model on sensitive customer data without transferring the raw data to a central server. Which privacy-preserving technique should they use?

Accepted Answer

Federated learning. Federated learning is the correct technique because it trains a shared model across decentralized edge devices holding local data, without transferring raw customer data to a central server. Only model updates (gradients) are sent to the aggregation server, preserving data locality and reducing exposure. This directly addresses the requirement of avoiding raw data transfer while still enabling collaborative model training.

Answer

Differential privacy

Answer

Data minimisation

Answer

Anonymisation

Question 6

An AI system is being deployed to detect deepfakes in video content. To comply with transparency obligations, what should the company implement?

Accepted Answer

A visible or invisible watermark on AI-generated videos. Transparency obligations under AI governance frameworks (e.g., EU AI Act) require clear disclosure when content is AI-generated. A visible or invisible watermark directly informs viewers that the video is synthetic, fulfilling this requirement without over-blocking legitimate content. Option B is correct because it provides a verifiable, non-disruptive method of labeling AI-generated media.

Answer

A system to automatically block all deepfake content

Answer

A process to report deepfake content to law enforcement

Answer

Encryption of the video files to prevent tampering

Question 7

An organization is implementing an AI governance framework. Which THREE components are essential for compliance with ethical AI standards?

Accepted Answer

Data privacy protection measures (e.g., differential privacy).. Option A (data privacy protection measures such as differential privacy) is essential because ethical AI standards require safeguarding personal data, and techniques like differential privacy provide formal, quantifiable guarantees that individual records cannot be inferred from model outputs, supporting regulations like GDPR. Option D (model explainability and interpretability mechanisms) is essential because stakeholders must understand how decisions are reached; methods such as SHAP, LIME, or inherently interpretable models enable accountability and contestability required by ethical frameworks. Option E (regular bias auditing of models) is essential because systematic auditing detects disparate impact across protected groups, using metrics like demographic parity or equalized odds, ensuring fairness is continuously monitored rather than assumed. Option B is not required because ethical compliance concerns how models are governed and used, not whether they are open-sourced; proprietary models can be fully ethical. Option C is incorrect because maximizing accuracy for revenue is a business objective, not an ethical compliance requirement, and can even conflict with fairness and privacy goals.

Answer

Open-source licensing of all models.

Answer

Maximizing model accuracy to increase revenue.

Question 8

Which THREE of the following are key components of an AI governance framework?

Accepted Answer

Regular auditing and monitoring for compliance.. A is correct because regular auditing and monitoring for compliance is a core governance control that provides ongoing assurance that AI systems adhere to policies, regulations, and internal standards, enabling detection and remediation of drift or violations. C is correct because ethical guidelines for AI development and deployment establish the principles (e.g., fairness, transparency, accountability, privacy) that direct how AI is designed and used, forming the normative backbone of any governance framework. D is correct because explainability mechanisms for model decisions support accountability and oversight by making model behavior interpretable to stakeholders, auditors, and regulators, which is essential for contestability and trust. B does not belong because cloud-based deployment for scalability is an infrastructure/architecture choice, not a governance component, and governance applies regardless of hosting model. E does not belong because model accuracy thresholds for production deployment are a performance/quality gate, not a governance framework component, and accuracy alone does not address compliance, ethics, or explainability.

Answer

Cloud-based deployment for scalability.

Answer

Model accuracy thresholds for production deployment.

Question 9

A retail company wants to build a model to predict customer churn based on purchase history and demographics. The dataset includes categorical features like region and gender, and numerical features like total spend. What is the best initial step before training the model?

Accepted Answer

One-hot encode categorical variables and normalize numerical variables. Categorical features like region and gender must be converted to numeric form via one-hot encoding (or similar) because most ML algorithms cannot consume raw strings. Numerical features like total spend should be normalized (e.g., min-max or z-score) so that features on different scales contribute equally during training. This preprocessing step is the correct initial action before model training.

Answer

Train a deep neural network directly on raw data

Answer

Remove all categorical features to simplify the model

Answer

Perform principal component analysis (PCA) on all features

Question 10

A team is considering whether to fine-tune a base LLM or use RAG for a question-answering system over a large, static corpus of scientific papers. The answer must be highly accurate and grounded in the papers. Which approach is BEST and why?

Accepted Answer

RAG because it retrieves and grounds answers in the source documents. RAG is the best fit because it retrieves relevant passages from the scientific corpus at query time and injects them into the LLM's context, grounding the answer in the actual source documents. This directly satisfies the requirement for high accuracy and traceability to the papers, and it avoids the cost and staleness issues of fine-tuning on a large static corpus. Fine-tuning changes model weights but does not guarantee the model will cite or stay faithful to specific documents.

Answer

Fine-tuning because it adapts the model to the scientific domain

Answer

Fine-tuning because it is faster at inference time

Answer

RAG because it does not require any labeled data

Question 11

A developer is implementing a RAG system and needs to choose a similarity metric for retrieving document chunks. The embedding model produces normalized vectors. Which metric is computationally efficient and equivalent to cosine similarity for normalized vectors?

Accepted Answer

Dot product. For normalized (unit-length) vectors, the dot product is mathematically equivalent to cosine similarity because the cosine formula divides by the product of magnitudes, which are both 1. Dot product is also computationally cheaper since it skips the normalization division, making it the efficient choice for RAG retrieval over normalized embeddings.

Answer

Euclidean distance

Answer

Hamming distance

Answer

Manhattan distance

Question 12

Which of the following best describes the difference between narrow AI and general AI?

Accepted Answer

Narrow AI is designed for a specific task; general AI aims to perform any cognitive task a human can.. Narrow AI specializes in one task, while general AI would possess human-like cognitive abilities across domains.

Answer

Narrow AI relies on supervised learning; general AI uses unsupervised learning exclusively.

Answer

Narrow AI requires large datasets; general AI can learn from few examples.

Answer

Narrow AI can perform any intellectual task; general AI is limited to specific tasks.

Question 13

An AI system uses a pre-trained image classification model to detect defects in manufacturing. The team wants to deploy the model in an edge device with limited GPU memory. Which technique should they consider first?

Accepted Answer

Apply quantization to reduce model size. Quantization reduces the precision of the model's weights and activations (e.g., from 32-bit floating point to 8-bit integer), which significantly shrinks the model size and memory footprint while often maintaining acceptable accuracy. This is the most direct and effective first step for deploying a pre-trained model on an edge device with limited GPU memory, as it requires no retraining and immediately addresses the memory constraint.

Answer

Train the model from scratch using a smaller dataset

Answer

Use a larger model with more parameters for higher accuracy

Answer

Increase the batch size to improve throughput

Question 14

A security researcher demonstrates that by adding small perturbations to an image of a stop sign, an autonomous vehicle's AI misclassifies it as a speed limit sign. This is an example of which type of attack?

Accepted Answer

Adversarial example attack. This is an adversarial example attack because the researcher adds imperceptible perturbations to the input image (the stop sign) to cause the AI model to output an incorrect classification (speed limit sign). Adversarial examples exploit the model's sensitivity to small, crafted changes in input data, leading to misclassification without altering the underlying task or training data.

Answer

Data poisoning attack

Answer

Model extraction attack

Answer

Membership inference attack

Question 15

An AI system for detecting anomalies in manufacturing sensor data uses a model trained on normal operation data only. During monitoring, the model flags many false positives. Which adjustment is MOST likely to reduce false positives?

Accepted Answer

Adjust the anomaly detection threshold to be less sensitive (e.g., require a higher reconstruction error). For an unsupervised anomaly detector trained only on normal data, false positives are typically controlled by tuning the decision threshold — e.g., requiring a higher reconstruction error (for autoencoders) or a lower anomaly score before flagging. Raising the threshold makes the model less sensitive, reducing false positives at the cost of potentially missing some true anomalies. This is the most direct, lowest-effort adjustment.

Answer

Switch from an autoencoder to a one-class SVM

Answer

Add synthetic anomalies to the training set and retrain as a supervised classifier

Answer

Increase the size of the training dataset with more normal operation data