OpenAI is dedicated to creating AI tools that possess the capability for self-optimization and has already initiated preparations to tackle the potential risks that may arise. Over the past half-year, remarkable strides have been made in the programming tools developed by OpenAI and Anthropic, with the possibility of 'recursive self-improvement' in AI systems emerging as a key focus within the industry. This technology empowers AI to contribute to the development of subsequent-generation models using its own abilities, thereby establishing a cycle of 'AI constructing AI' that can perpetually iterate without human intervention. However, this process is not without its challenges, including immense computational power requirements, the risk of cognitive closed loops, and the unpredictability of losing control. To counteract these risks, OpenAI is actively seeking out security researchers to delve into the potential repercussions of AI self-evolution, devise tools to safeguard models against data poisoning attacks, and keep tabs on the advancement of technological automation.