A security analyst notices that regardless of user-submitted prompts, an AI model always returns unsanitized responses. These responses are then passed to multi...