Adversarial Examples: Tricking AI Models with Subtle Inputs
Adversarial examples are intentionally crafted inputs that appear benign to humans but are designed to trick AI models into making incorrect predictions or misclassifications, exploiting sensitivities in their decision boundaries.