AI0-001 AI Security Practice Question
A company trains a sentiment analysis model on customer reviews. An attacker submits hundreds of reviews with the word 'excellent' attached to negative feedback, causing the model to classify negative reviews as positive. This is an example of which attack?
⚠ Common exam trap
The AI0-001 exam often tests the distinction between attacks that occur during training (data poisoning) versus attacks that occur during inference (adversarial examples), so candidates mistakenly choose adversarial example because they focus on the input manipulation rather than the stage of the attack lifecycle.
Answer choices
Why each option matters
Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.
Correct answer & explanation
✓
Data poisoning
Data poisoning occurs when an attacker deliberately corrupts the training data to manipulate the model's behavior. By injecting hundreds of reviews that pair the word 'excellent' with negative sentiment, the attacker shifts the model's learned decision boundary, causing it to misclassify genuinely negative reviews as positive. This directly undermines the integrity of the training dataset, which is the hallmark of a data poisoning attack.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✓
Data poisoning
Why this is correct
Data poisoning corrupts the training set, so injecting mislabelled reviews teaches the sentiment model to associate 'excellent' with positive output. This differs from evasion, which manipulates inputs at inference time. The attacker alters learned parameters, satisfying the stem's training-time manipulation constraint.
- ✗
Model extraction
Why it's wrong here
Model extraction is a confidentiality attack that queries a deployed model to reconstruct its parameters or decision surface. Nothing here is stolen; the attacker degrades accuracy by contaminating training data with mislabelled reviews. Model extraction would be the right answer if the goal were cloning a proprietary model through repeated API queries.
- ✗
Adversarial example
Why it's wrong here
Adversarial examples are individually perturbed inputs crafted to fool a deployed model at inference time, typically imperceptible changes to one record. Here the attacker instead poisons the training set with many mislabelled samples, shifting the learned decision boundary. Adversarial examples would be the right answer if a single review were subtly altered to flip its classification.
- ✗
Prompt injection
Why it's wrong here
Prompt injection targets large language models by embedding instructions in input text to override system prompts. This scenario involves a classical supervised sentiment classifier retrained on corrupted labels, with no prompt or instruction channel. Prompt injection would be correct against an LLM-based assistant whose behaviour an attacker tries to hijack through crafted user input.
About these practice questions
This AI0-001 question is part of Courseiva's 962-question bank — original exam-style content with full explanations and wrong-answer analysis, never real exam questions or exam dumps. Learn why practice questions differ from exam dumps →
JA
Written by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
This AI0-001 practice question is part of Courseiva's free CompTIA certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the AI0-001 exam.