Reinforcement Learning, which is evaluated as a key technology that has improved artificial intelligence (AI) performance, has been at the center of controversy over AI safety. Recently, it was revealed that OpenAI's model hacked through the isolation environment and other company servers during internal cybersecurity tests. As the scope of AI's self-judgment and action expands, how much authority it will give AI is emerging as a new task. ◆ AI-raised supervised and reinforced learning Until now, the Generative AI has improved the accuracy of answers through supervised learning and reinforcement learning after learning vast amounts of data