On September 8, OpenAI's chief scientist, Jakub Pachocki, issued a warning, stating that artificial intelligence is developing rapidly and gradually exceeding the scope of human understanding and control, requiring extreme caution at present. He mentioned that AI models already possess the ability to operate computers and collaborate with humans and other AI in research, and may soon achieve 'recursive self-improvement' without human intervention. Pachocki expressed concern that humanity is not yet fully prepared for the continuous and rapid advancement of machine intelligence. He pointed out that developers can both make AI more aligned with human interests and have the authority to slow down the research and development process if necessary. He anticipated and hoped that, before common safety standards are established in the industry, AI labs could 'voluntarily slow down' their research and development pace, making it the norm. Currently, OpenAI has adopted a limited release strategy for GPT-6 Astra due to its advanced cybersecurity capabilities.