SoftBank CEO Masayoshi Son told CNBC on June 5 that OpenAI’s next frontier model is being designed by AI rather than by human engineers—a claim he said he heard directly from OpenAI CEO Sam Altman and company engineers. Son described the development as a clear sign that artificial superintelligence (ASI) is arriving far sooner than previously expected, asserting that current AI models are already “smarter than humans in most subjects” and that models designed by AI will surpass human intelligence by an exponential margin. He revised his earlier ten-year ASI timeline down to roughly two years.
Son’s comments reflect SoftBank’s deepening all-in bet on OpenAI, into which the group has committed tens of billions of dollars, including through the Stargate AI infrastructure project. His remarks arrive as AI safety researchers, including those at Anthropic, have raised concern that recursive self-improvement—AI autonomously designing increasingly capable successors—may be approaching faster than institutions are prepared to govern. OpenAI did not publicly comment on unreleased models in response to Son’s statements.