Question 1

An attacker sends carefully crafted queries to a language model with the goal of reconstructing the training dataset, including sensitive records. Which threat does this represent?

Options :

A : Input fuzzing

B : Model inversion attack

C : Gradient leakage

D : Adversarial sampling

Answer: B

Question 2

A dataset contains a feature with 100 categories, but only 5 samples are available for each category. What potential issue could arise when training a model on this dataset?

Options :

A : Data imbalance

B : Data normalization

C : Underfitting

D : Overfitting

Answer: A

Question 3

A multinational company deploys a customer service chatbot that frequently fails to respond accurately in non-English languages. What AI adoption challenges are involved in this scenario?

Options :

A : Scalability issues

B : Data quality and preprocessing

C : Language-specific training limitations

D : GPU memory limits

Answer: B,C

Question 4

An AI chatbot is designed without sanitizing user inputs before sending them to an intent classification model. What risk does this pose to the system?

Options :

A : Model compression failure

B : Prompt injection or logic manipulation

C : Increased false negatives

D : Loss of training data

Answer: B

Question 5

An insurance company rolls out a claims processing AI tool across regional offices. In some regions, staff find the tool ineffective due to varying policy rules and data entry standards. Which AI adoption challenges are evident here?

Options :

A : Limited bandwidth

B : Lack of localization

C : Input heterogeneity

D : Ambiguous labeling

Answer: B,C

Free AI-Essentials Mock Exam – Practice Online Confidently

Buying Options: