ISACA AI Fundamentals Practice Questions with Explanations

Free ISACA AI Fundamentals practice questions. 22 of them, each with the correct answer, a full explanation, and the reason every other option is wrong.

AI Fundamentals practice questions

Question 1

Which of the following methods scales poorly with large data sets?

  • A. Fuzzy
  • B. Density-based
  • C. Hierarchical
Show answer and explanation ▾

Correct answer: C

Hierarchical clustering compares every pair of items, so its time and memory needs grow quickly as the data set grows. That makes it a poor fit for large data.

Why the other options are wrong:

  • A. Fuzzy clustering costs more than simple methods but is not the one known for scaling badly here.
  • B. Density-based methods such as DBSCAN handle large sets reasonably well, especially with an index.

Question 2

Which of the following trains a predictive model from input data?

  • A. Expert system
  • B. Artificial intelligence
  • C. Machine learning
Show answer and explanation ▾

Correct answer: C

Machine learning fits a model to example input data so that it can predict outcomes for new data.

Why the other options are wrong:

  • A. An expert system applies rules written by people rather than learning a model from data.
  • B. Artificial intelligence is the broad field; training a model from data is the job of machine learning within it.

Question 3

Which of the following ethical challenges is the use of AI MOST likely to present?

  • A. Bias in content output
  • B. Unauthorized alteration of data
  • C. Data misuse
Show answer and explanation ▾

Correct answer: A

AI models learn from data that can carry bias, so their output can be unfair or skewed. That risk comes from how AI works, which is why it is the ethical challenge most specific to AI.

Why the other options are wrong:

  • B. Unauthorized alteration of data is a general security problem, not specific to AI.
  • C. Data misuse can happen with any system and is a data governance problem more than an AI-specific one.

Question 4

Which of the following is the MOST significant outcome from flawed program code that is likely to lead to ethical concerns?

  • A. Additional training requirements for developers
  • B. Increased costs of writing new code
  • C. Greater unpredictability of results
Show answer and explanation ▾

Correct answer: C

Flawed code makes a system behave in ways nobody intended, so outcomes become harder to predict. Unpredictable results are what create ethical concerns such as unfair or harmful decisions.

Why the other options are wrong:

  • A. More developer training is a business cost, not an ethical concern.
  • B. Higher coding costs are a financial effect, not an ethical one.

Question 5

Which of the following logic types is BEST to use for an AI algorithm that determines whether drivers are stopping for red traffic lights?

  • A. Spatiotemporal
  • B. Temporal
  • C. Boolean
Show answer and explanation ▾

Correct answer: B

Temporal logic reasons about conditions over time, such as whether a vehicle stops while a light is red. The task depends on the timing of the light state and the driver's action.

Why the other options are wrong:

  • A. Spatiotemporal logic adds location reasoning to time, which is more than this task needs.
  • C. Boolean logic only handles true or false values and does not capture behavior over time.

Question 6

Which of the following is an example of video synthesis?

  • A. Object resolution
  • B. Face-swapping
  • C. Image classification
Show answer and explanation ▾

Correct answer: B

Face-swapping generates new video by replacing one person's face with another's, which is a form of video synthesis. Deepfakes are a common example.

Why the other options are wrong:

  • A. Object resolution is not an example of generating new video content.
  • C. Classifying pictures assigns labels to existing pictures and does not synthesize video.

Question 7

Which type of data is MOST commonly augmented using techniques like pitch shifting and time stretching?

  • A. Text
  • B. Audio
  • C. Images
Show answer and explanation ▾

Correct answer: B

Pitch shifting and time stretching change the sound of a recording while keeping its meaning, so they are standard augmentation techniques for audio data.

Why the other options are wrong:

  • A. Text is augmented with methods such as synonym replacement, not pitch or timing changes.
  • C. Pictures are augmented with methods such as rotation, cropping and flipping.

Question 8

The R2 notation can range from 0-100%. An R2 value of 0.845 signifies which type of a relationship?

  • A. Neutral association
  • B. Positive linear association
  • C. Negative association
Show answer and explanation ▾

Correct answer: B

R2 is the share of the variation in the outcome that a linear model explains, so a value of 0.845, about 84.5 percent, shows a strong linear association between the variables. A neutral association would give a value near zero.

Why the other options are wrong:

  • A. A neutral association would give an R2 near zero, not a value this high.
  • C. R2 is a squared measure that is never negative, so the value itself does not point to a negative association.

Question 9

Which of the following is an example of the use of AI to assist consumers in finding new or old relationship connections?

  • A. Apps that use facial recognition
  • B. Apps that use a global positioning system (GPS)
  • C. Apps that look at possible connection proximity
Show answer and explanation ▾

Correct answer: A

Facial recognition lets apps match faces in photos to people, which helps consumers find new or past connections. It is an AI technique applied to photos.

Why the other options are wrong:

  • B. GPS provides location data and is not itself an AI technique for finding connections.
  • C. Checking the proximity of possible connections relies on location, not on AI recognition.

Question 10

Which of the following expert system components retrieves the information when a user queries the system?

  • A. Knowledge base
  • B. Inference engine
  • C. Explanation module
Show answer and explanation ▾

Correct answer: B

The inference engine applies rules to the stored facts to find and derive the information that answers a user's query.

Why the other options are wrong:

  • A. The knowledge base stores facts and rules but does not itself retrieve and reason over them.
  • C. The explanation module shows how a conclusion was reached, not how information is retrieved.

Question 11

Leveraging AI, data processing to scan systems for threats and weaknesses supports which of the following functions?

  • A. Cybersecurity
  • B. Superintelligence
  • C. Marketing
Show answer and explanation ▾

Correct answer: A

Using AI to process data and scan systems for threats and weaknesses supports the cybersecurity function.

Why the other options are wrong:

  • B. Superintelligence is a hypothetical level of AI, not a business function.
  • C. Marketing uses AI for customer analysis and targeting, not for scanning systems for threats.

Question 12

Which of the following is a type of probability sampling?

  • A. Sampling fraction
  • B. Cluster
  • C. Judgment
Show answer and explanation ▾

Correct answer: B

Cluster sampling divides the population into groups and randomly selects whole groups, so every unit has a known chance of selection. That makes it a probability sampling method.

Why the other options are wrong:

  • A. A sampling fraction is a ratio of sample size to population size, not a sampling type.
  • C. Judgment sampling relies on the researcher's choice, which makes it non-probability sampling.

Question 13

Which of the following can be used by organizations to detect possible anomalies in employee actions?

  • A. Deep neural networks
  • B. Confusion matrix
  • C. Behavioral AI
Show answer and explanation ▾

Correct answer: C

Behavioral AI learns normal patterns of user activity and flags deviations, so organizations use it to detect possible anomalies in employee actions.

Why the other options are wrong:

  • A. Deep neural networks are a general model type, not a tool specific to monitoring employee behavior.
  • B. A confusion matrix evaluates classification accuracy and does not detect anomalies.

Question 14

Combining AI with an individual's digital footprint can pose a significant risk due to which of the following algorithms?

  • A. Cryptoeconomics
  • B. Generative adversarial networks
  • C. Watson
Show answer and explanation ▾

Correct answer: B

Generative adversarial networks can use personal data from a digital footprint to produce realistic fake content, such as deepfakes, which creates significant risk.

Why the other options are wrong:

  • A. Cryptoeconomics concerns incentive design in blockchain systems, not generating content from personal data.
  • C. Watson is a specific IBM AI platform and not the algorithm behind this risk.

Question 15

Classification models are a form of which model type?

  • A. Supervised learning
  • B. Reinforcement learning
  • C. Unsupervised learning
Show answer and explanation ▾

Correct answer: A

Classification models are trained on labeled examples to predict categories, which makes them a form of supervised learning.

Why the other options are wrong:

  • B. Reinforcement learning learns from rewards through interaction, not from labeled examples.
  • C. Unsupervised learning works on unlabeled data and finds structure, such as clusters.

Question 16

Which of the following is an example of weaponizing AI?

  • A. Preventing individuals from driving without a driver's license
  • B. Preventing individuals from reproducing due to certain genetic markers
  • C. Using electric smart meters
Show answer and explanation ▾

Correct answer: B

Using AI to prevent people from reproducing based on genetic markers turns the technology into a tool of control and harm against individuals, which is weaponization.

Why the other options are wrong:

  • A. Preventing unlicensed driving is an ordinary regulatory control, not weaponization.
  • C. Electric smart meters are utility devices and are not a weaponized use of AI.

Question 17

Which of the following is of GREATEST concern given that AI's ethical framework is still in its infancy?

  • A. Financial responsibility
  • B. Accountability
  • C. Intellectual property
Show answer and explanation ▾

Correct answer: B

With ethical frameworks for AI still immature, it is unclear who answers for harmful decisions made by AI systems, so accountability is the greatest concern.

Why the other options are wrong:

  • A. Financial responsibility is a narrower issue that depends on first settling who is accountable.
  • C. Intellectual property is a legal concern that is not the greatest ethical gap.

Question 18

Which of the following is an example of a job that can now be performed by an AI?

  • A. Valet parking
  • B. Online Instruction
  • C. Commercial flying
Show answer and explanation ▾

Correct answer: B

AI systems can deliver online instruction, such as tutoring and course delivery, so it is a job AI can now perform.

Why the other options are wrong:

  • A. Valet parking requires physical handling of vehicles in varied environments.
  • C. Commercial flying requires a certified human pilot in command.

Question 19

Which of the following is the MOST common reason for enterprises to incorporate AI in order to achieve a competitive advantage?

  • A. Cost reductions
  • B. Risk remediation
  • C. Targeted marketing
Show answer and explanation ▾

Correct answer: A

Enterprises most commonly adopt AI to automate work and improve efficiency, which reduces costs and gives a competitive advantage.

Why the other options are wrong:

  • B. Risk remediation is not the most common reason enterprises adopt AI for competitive advantage.
  • C. Targeted marketing is one use case and is less common than cost reduction as a driver.

Question 20

A clustering method is considered good when it produces which of the following?

  • A. Partitions
  • B. Intra-classes
  • C. Inter-classes
Show answer and explanation ▾

Correct answer: B

A good clustering method produces clusters whose members are highly similar to each other, which is high intra-class similarity.

Why the other options are wrong:

  • A. Partitions describe how data is divided and do not measure clustering quality.
  • C. A good clustering aims for low similarity between classes, so inter-class similarity is not what it seeks to produce.

Question 21

Which of the following expertise domains can be used to simulate crimes and detect patterns?

  • A. Business expertise
  • B. Forensic science
  • C. Human science
Show answer and explanation ▾

Correct answer: B

Forensic science uses modeling and pattern analysis to simulate crimes and detect patterns in evidence and behavior.

Why the other options are wrong:

  • A. Business expertise applies to commercial operations, not crime simulation.
  • C. Human science is a broad field and is not the domain that simulates crimes.

Question 22

Machine learning can be defined as:

  • A. a predetermined set of rules.
  • B. a type of automation.
  • C. a predictive model.
Show answer and explanation ▾

Correct answer: C

Machine learning builds models from data to make predictions, so it can be defined as a predictive model.

Why the other options are wrong:

  • A. A predetermined set of rules describes traditional programming, not learning from data.
  • B. A type of automation is too broad and does not capture learning from data.

That was 22 of 60.

The full ISACA AI Fundamentals pack has all 60 questions, each with the answer, the explanation and why the other options are wrong, plus a questions-only copy for timed runs. $39.00 USD, paid once, with free monthly updates and a pass-or-your-money-back guarantee.

Get the full pack

Not ready yet? Get the first 22 free as one PDF by email. Get the Half Pack →

About this exam

These 22 ISACA AI Fundamentals questions are free here, and as a free PDF by email.

They are drawn from the same bank as the full AI Fundamentals pack, which has 60 questions in total.

Back to blog