AI is designing OpenAI's next model, says SoftBank CEO amid Anthropic's call for AI slowdown

SoftBank CEO Masayoshi Son revealed in a CNBC interview that OpenAI engineers, including CEO Sam Altman, told him an AI model is already helping design the company’s next-generation system. This claim aligns with concerns raised by Anthropic, which has urged governments and AI firms to prepare for a future where AI systems could autonomously design, build, and train successors with minimal human input. Son described this as the beginning of 'superintelligence,' where AI surpasses human capabilities and increasingly drives its own evolution. He stated that engineers may soon lack the expertise to outpace AI-driven development, with models eventually generating their own successors exponentially faster than human-led progress. Anthropic has argued for a regulatory 'brake pedal' to slow advanced AI development if it outpaces societal control, a stance echoed by co-founder Jack Clark. OpenAI, while declining to comment on unreleased models, previously confirmed AI-assisted development in its GPT-5.3-Codex model, which used early iterations to debug training and deployment processes. Son’s predictions have grown more aggressive, now suggesting superintelligence—defined as AI 10,000 times smarter than humans—could emerge within two years, down from his earlier 10-year estimate. The shift reflects accelerating AI capabilities, though industry experts remain divided on timelines and risks.