With the rapid advancement of AI utilization, new attack methods targeting AI systems are emerging one after another. This presentation offers a comprehensive overview of a wide range of attack techniques and their impacts on AI systems—starting with evasion and poisoning attacks that have been known since the early days of AI security research and extending to more recent threats such as prompt injection and code injection into AI models. The talk will systematically organize the impacts of these attacks on AI systems in line with the CIA security triad (confidentiality, integrity, and availability) and present intuitive visualization models to facilitate a deeper understanding of the attacks.