What Is Adversarial Example?

Adversarial Example — an input deliberately crafted with small, often imperceptible changes that cause an AI model to make a confident but wrong prediction.

Adversarial examples exploit the gap between how a model "sees" data and how a human does — a tiny, targeted perturbation can flip a classification while looking unchanged to a person. They are a core concern for the robustness and security of high-stakes AI, and a reason adversarial testing is part of serious model evaluation.

Source: Machine-learning security research