This study investigated the dynamics of user interactions with AI agents, specififically delving into the impact of conversation types that users hold with an AI agent on self-disclosure and privacy insensitivity toward AI agents. The present study also examined the interplay of conversation types with attitudes toward the machine (i.e., perceived humanness and intimacy perception). Results exhibited that both functional and emotional conversations with AI agents were signifificantly associated with self-disclosure to AI agents. The more functional or emotional conversation a user made with AI agents, the more likely the user was to disclose his/her information to the devices. And, the impact of emotional conversation was found to be significantly greater than that of functional conversation. Yet, only emotional conversation was associated with AI privacy insensitivity. The more a user made emotional conversations with AI agents, the more likely the user was insensitive to privacy issues related to AI agents. Moreover, perceived humanness played a role in strengthening the relationship between functional conversation and self-disclosure, whereas emotional conversations with AI agents were more positively related to AI privacy insensitivity when users perceived the agents as human-like. Discussion and limitations were further addressed.