ISACA AI Fundamentals

The ISACA AI Fundamentals is an entry-level certificate designed to provide a foundational understanding of artificial intelligence, its business applications, and the associated risks. Unlike ISACA's professional certifications (like CISA or CISM), this is a knowledge-based certificate that does not require prior experience or a rigorous application process

---------- Question 1

A customer service chatbot that effectively interprets varied natural language queries, understands user intent, and generates relevant, coherent responses is a prime application of which specialized machine learning subfield?

1. Computer Vision, for processing visual inputs.
2. Robotic Process Automation, for automating repetitive tasks.
3. Natural Language Processing, for understanding and generating human language.
4. Reinforcement Learning, for learning optimal actions through rewards.

---------- Question 2

Data governance plays a critical role in AI project success. Which aspect of data governance directly addresses the ethical use of personal identifiable information PII within AI systems?

1. Data storage optimization and backup strategies
2. Ensuring data quality and integrity for model training
3. Establishing policies for data privacy, consent, and anonymization or pseudonymization
4. Monitoring infrastructure uptime and network security

---------- Question 3

An AI system developed for loan approvals consistently rejects applications from a specific demographic group, even when creditworthiness is comparable to approved applicants. This scenario highlights a significant concern related to which AI risk?

1. Overfitting
2. Data leakage
3. Algorithmic bias
4. Model explainability

---------- Question 4

The majority of AI systems successfully deployed today are characterized as narrow AI. What is the defining characteristic of narrow AI?

1. It possesses consciousness and self-awareness
2. It can perform any intellectual task a human can
3. It specializes in a single task or a limited set of tasks
4. It continuously learns and adapts without human intervention

---------- Question 5

To train a robust machine learning model for detecting fraudulent transactions, it is crucial that the training dataset accurately mirrors the distribution and characteristics of real-world transactions across all customer segments. This best relates to which statistical concept?

1. Outlier detection.
2. Sample representativeness.
3. Covariance analysis.
4. Hypothesis testing.

---------- Question 6

What is the primary ethical and legal concern that arises when an AI system is deployed to make high-stakes decisions that directly impact individuals, such as creditworthiness assessments or predictive policing?

1. The system's raw computational efficiency and speed of processing.
2. The potential for algorithmic bias, lack of transparency, and unfair outcomes.
3. The financial cost associated with storing the vast amounts of training data.
4. The technical requirement for real-time processing capabilities.

---------- Question 7

A large e-commerce company wants to identify distinct customer segments based on their purchasing history, browsing behavior, and demographic information without any predefined categories. Which machine learning approach is most appropriate for this task?

1. Supervised classification.
2. Reinforcement learning.
3. Unsupervised clustering.
4. Regression analysis.

---------- Question 8

A malicious actor attempts to subtly modify input data to an image recognition AI, causing it to misclassify an object (e.g., classifying a stop sign as a yield sign), even though the changes are imperceptible to humans. What type of AI security threat does this scenario represent?

1. Data poisoning, where training data is corrupted.
2. Model inversion, reconstructing training data from model outputs.
3. Adversarial attack, manipulating input to cause misclassification.
4. Model theft, unauthorized copying of a proprietary AI model.

---------- Question 9

During the deployment of an AI model into a production environment, which security practice is paramount to protect the model from adversarial attacks and ensure data integrity?

1. Granting all users full administrative access to the model server.
2. Disabling all network firewalls to ensure maximum model availability.
3. Implementing robust access controls, encryption, and continuous monitoring for anomalies.
4. Storing model weights and training data on publicly accessible cloud storage.

---------- Question 10

Why is population sampling a critical consideration when preparing data for AI model training?

1. To ensure the training data is perfectly identical to the entire population to avoid any data loss.
2. To significantly reduce the computational resources required for model training and deployment.
3. To create a representative dataset that minimizes bias and accurately reflects the underlying distribution of the broader population.
4. To facilitate the encryption of sensitive information within the dataset for enhanced security.

---

Are they useful?

Click here to get 360 more questions to pass this certification at the first try! Explanation for each option is included!

Follow the below LINKEDIN channel to stay updated about 89+ exams!

AI Strategy and Engineering Hub