Adversarial attacks are manipulated inputs that can move models towards potential system failures with serious consequences in AI fields like autonomous driving, finance, healthcare, and security. This paper offers a comprehensive review of adversarial attack strategies, examining how these attacks exploit AI vulnerabilities and how they differ across model architectures. By systematically identifying and classifying various attack types, we highlight specific weaknesses in models such as Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs) and Graph Neural Networks (GNNs). We also discuss defense mechanisms and recovery strategies, highlighting the importance of Responsible AI (RAI) principles to foster ethical deployment, fairness, transparency and accountability in AI systems. This exploration aims to enhance our understanding of adversarial threats guiding future research to strengthen AI resilience in real-world applications.