Masayoshi Son, the founder of SoftBank Group Corp, has long been one of the most vocal champions of artificial intelligence, yet the Japanese billionaire has now issued a rare warning — as AI capabilities expand at breakneck speed, superintelligence could become "extremely dangerous" if it falls into the wrong hands.
Speaking on Sunday at an event on the sidelines of the STS Forum in Kyoto, Son said the exponential leap in AI capabilities makes it imperative for nations to build mutual trust and work together to steer the technology. He warned that humanity has "no room left" for countries to confront one another, because the threat posed by powerful AI models is far more severe than conflicts among humans.
The remarks stand in sharp contrast to Son's consistently optimistic tone, reflecting growing anxiety across the industry after a string of unsettling safety failures in AI models in recent months. For investors, the signal is that AI safety is moving from a fringe discussion to a core concern, and pressure for regulation and international coordination is likely to intensify further.
The Rare Warning: Safety Failures Spark Concern
Now 69 years old, Son is one of the most prominent backers in the global AI sector.
Funds under SoftBank Group Corp have invested a cumulative nearly $65 billion in ChatGPT developer OpenAI, and Son has for years publicly touted AI's potential to deliver investment returns and benefit humanity.
Because of this, his latest comments are seen by outsiders as an "atypical" cautious signal. A string of unsettling AI model safety failures in recent months is believed to be the direct trigger that prompted Son's warning. In his remarks, he made clear that countries must cooperate to address the greater threat posed by powerful AI models, saying "superintelligence in the wrong hands could become super dangerous."
US-Led 17-Nation Coalition: Accelerating Scientific Research with AI
Also speaking alongside Son was Michael Kratsios, a White House science and technology policy advisor. He made a special trip to Kyoto to announce a US-led initiative involving 17 countries aimed at better harnessing AI to accelerate scientific research. Representatives from Japan, South Korea, the United Kingdom, Germany, Singapore, the United Arab Emirates and other nations took the stage to take part in the event.
Kratsios painted a picture of AI-driven research: in automated cloud laboratories operating at industrial scale, robotic systems would run hundreds of parallel experiments simultaneously, generating data in real time and feeding it back for analysis and hypothesis refinement.
He also said AI is already capable of synthesizing thousands of papers, compressing years of research into "tens of hours of work by human researchers with the help of a digital army of agents."
A Three-Stage Evolution Path: Entering the Physical World in 2026
In his speech, Son reiterated his judgment on the trajectory of AI development.
He said the capabilities of AI services will continue to grow exponentially: the technology currently powering chatbots is already helping AI services browse the web, write code and consume ever more general-purpose chip compute; in the next phase, related programs will enter a third stage starting next year, ready to help robots and other devices interact with the physical world.
On the longer-term outlook, Son restated his earlier forecast: by 2040, AI and machines capable of learning will account for at least 20% of global GDP, equivalent to $46 trillion. "This is just the beginning," he said.