Adversarial Attack

Like a targeted trick designed to intentionally fool an AI model. These are crafted inputs that look normal to humans but cause the model to make a wrong prediction or generate harmful output. Think of it as disguising a stop sign with subtle stickers that a human can easily ignore, but an autonomous vehicle might misinterpret as a speed limit sign. It highlights a critical vulnerability in AI systems, showing that even high-performing models can be manipulated.