An attack on AI models involving inputs intentionally designed to confuse or mislead the model, causing it to make incorrect predictions or classifications