A company deploys an AI model to screen job applications. The model is trained on historical hiring data, which reflects past biases. As a result, the model systematically disadvantages certain demographic groups. What is the primary limitation of AI systems that this scenario illustrates?

The model lacks sufficient computational power

Bias in training data leads to biased model outcomes

The model is overfitting to the training data

The model is not using deep learning

Why do many AI models struggle with 'adversarial examples'?

Because they rely on patterns that can be subtly manipulated to cause misclassification.

Because they are too large to run on standard hardware.

Because they require too much labeled data.

Because they cannot process images.

What is a common limitation of AI systems that are trained on historical data?

They can only work with structured data.

They do not require any data to operate.

They are always 100% accurate on new data.

They can perpetuate or amplify biases present in the training data.

A data science team is building an image recognition system to detect suspicious objects in security camera footage. They have millions of labeled images and need a model that can automatically learn hierarchical features, from edges to object parts. Which AI subfield is best suited for this task?

Classical machine learning with decision trees

A research team is working on a system that can generate realistic human faces from random noise. They plan to use a neural network architecture that consists of two networks: one generates images and the other evaluates them. Which AI application and subfield does this represent?

Natural language processing with a transformer

Generative adversarial networks (GANs), a deep learning application

Reinforcement learning with a reward model

Classical machine learning with a decision tree

GIAC Offensive AI Analyst

Domain 1Objective 1

Artificial Intelligence Fundamentals GOAA Practice Questions (Page 7)

Part of the AI Foundations and Social Engineering domain, which makes up ~29% of our current practice bank. GIAC (SANS) does not publish an official question count, but from its 120-minute exam (~50–80 total, ~15–23 in this domain), expect 5–8 from this objective — we provide 48 practice questions to prepare you well beyond it. (estimate)

48questions here

10free pages

6concepts

Questions 31–35

31expert · hard
A company deploys an AI model to screen job applications. The model is trained on historical hiring data, which reflects past biases. As a result, the model systematically disadvantages certain demographic groups. What is the primary limitation of AI systems that this scenario illustrates?
Select an answer first
32foundation · easy
Why do many AI models struggle with 'adversarial examples'?
Select an answer first
33foundation · easy
What is a common limitation of AI systems that are trained on historical data?
Select an answer first
34application · medium
A data science team is building an image recognition system to detect suspicious objects in security camera footage. They have millions of labeled images and need a model that can automatically learn hierarchical features, from edges to object parts. Which AI subfield is best suited for this task?
Select an answer first
35application · medium
A research team is working on a system that can generate realistic human faces from random noise. They plan to use a neural network architecture that consists of two networks: one generates images and the other evaluates them. Which AI application and subfield does this represent?
Select an answer first

Finished these 5 questions?

Review the revealed explanations, or continue through the curriculum.

Next 5 questions →Back to GOAA

Free Basic Practice is a study aid with revealable answers — not a scored exam. Examers.io is independent and not affiliated with or endorsed by GIAC (SANS). “GOAA” is a trademark of its owner, used for identification only.