SoftBank's Masayoshi Son warns superintelligence in the wrong hands could be super dangerous

SoftBank Chief Sounds Alarm on AI Safety

Masayoshi Son, the billionaire chief of SoftBank Group Corp who has invested $64.6bn into OpenAI, delivered an uncharacteristic warning about AI safety at the Science and Technology in Society forum in Kyoto on Sunday1

1

2

2

. Speaking alongside Michael Kratsios, the White House science and tech policy advisor, Son stated that superintelligence in the wrong hands could become "super dangerous"1

1

. The 69-year-old Japanese billionaire emphasized that countries must build trust and work together to control threats from powerful AI models, declaring "we don't have any more luxury of having human fight against human"2

2

.

Massive Investment Despite Safety Concerns

SoftBank's Masayoshi Son has been directing substantial funds into OpenAI since 2024, with the company finishing a $30bn tranche this month that brought its total investment to $64.6bn, representing roughly 13% of OpenAI1

1

. The final $10bn came from the largest high-yield corporate bond sale on record, raised a fortnight ago1

1

. Despite his cautionary note on AI safety risks, Son has spent the past year calling bubble talk absurd and arguing that the technology would repay the investment1

1

. His warning reflects escalating concern following disturbing security breaches by AI models in recent months2

2

.

Kyoto Vision Declaration Gains International Support

Michael Kratsios arrived in Kyoto to launch the Kyoto Vision for a Golden Age of Science, a US-led initiative endorsed by 17 countries aimed at advancing scientific research using AI1

1

2

2

. The declaration envisions robotic laboratories running hundreds of parallel experiments simultaneously, with machines synthesizing thousands of research papers and compressing years of research into tens of hours of work by digital agents assisting human researchers2

2

. Seven European nations endorsed the vision: Germany, Italy, Poland, Greece, Bulgaria, Cyprus, and the United Kingdom1

1

. Six of these are EU member states bound by the Artificial Intelligence Act, whose labeling rules began applying in August1

1

. Other endorsers include Japan, South Korea, Singapore, the United Arab Emirates, Argentina, Chile, Indonesia, Kazakhstan, and New Zealand1

1

2

2

. Notably, China was not among the countries that initially endorsed the White House statement2

2

.

Economic Forecast and Physical World Integration

Son reiterated his projection that AI and machines with learning capabilities would account for at least 20% of global gross domestic product by 2040, equivalent to $46tn1

1

2

2

. He outlined the evolution of AI capabilities, noting that chatbots currently help services browse sites and write code using rising amounts of general-purpose chips2

2

. Son predicted that machines acting in the physical world would begin next year, marking a third phase of AI development1

1

. Supporting this vision, SoftBank led a $200M funding round in Zurich-based Gravis Robotics in August and has been in discussions with Munich's Agile Robots1

1

. The emphasis on global cooperation to control AI threats comes as nations grapple with balancing innovation against potential dangers from increasingly capable systems, making international cooperation to manage AI dangers essential for safely harnessing superintelligence benefits while mitigating risks.