Answer
Data poisoning attacks involve intentionally inserting harmful or misleading data into a training dataset to corrupt how an AI model learns.
Real-world example An attacker might poison a spam filter's training data so that certain harmful emails are incorrectly learned as safe.