Sam Altman, CEO of OpenAI: We have suspended some cutting-edge model reinforcement learning training to ensure that we can meet the corresponding alignment, security, and monitoring standards for the current new level of competency. Model iteration is developing rapidly now, and we have stated before that once we believe that the development of model capabilities exceeds the speed of safety and alignment work, we will take action. We take AI security very seriously. We believe the entire industry must collaborate on common safety standards, but in the meantime, we will also act unilaterally. We anticipate that security confidence will increasingly determine the pace of AI development. While we are optimistic about the ongoing alignment work, we remain committed to making cutting-edge technology capabilities widely used.

Sam Altman, CEO of OpenAI: We have suspended some cutting-edge model reinforcement learning training to ensure that we can meet the corresponding alignment, security, and monitoring standards for the current new level of competency. Model iteration is developing rapidly now, and we have stated before that once we believe that the development of model capabilities exceeds the speed of safety and alignment work, we will take action. We take AI security very seriously. We believe the entire industry must collaborate on common safety standards, but in the meantime, we will also act unilaterally. We anticipate that security confidence will increasingly determine the pace of AI development. While we are optimistic about the ongoing alignment work, we remain committed to making cutting-edge technology capabilities widely used.