SoftBank’s Masayoshi Son says superintelligence in the wrong hands could be “super dangerous”

* Masayoshi Son put $64.6 billion into OpenAI.
* Son stated that superintelligence in the wrong hands could become "super dangerous."
* Son called for countries to cooperate to control superintelligence.
* Son spoke beside Trump’s science adviser in Kyoto.
* SoftBank paid a $30 billion tranche this month, bringing its total investment in OpenAI to $64.6 billion.
* Michael Kratsios, who runs the White House science office, arrived with a plan for pointing the technology at scientific research.
* Seventeen countries endorsed a Kyoto Vision for a Golden Age of Science involving robotic laboratories.
* Germany, Italy, Poland, Greece, Bulgaria, Cyprus, and the UK endorsed the vision.
* China was not among the endorsers; Argentina, Chile, Indonesia, Kazakhstan, and New Zealand were endorsers.
* Son forecasted AI would be a fifth of global output by 2040, which he now estimates at $46 trillion.
* The declaration itself notes that countries endeavor to act but does not pledge or commit.
* The next phase, machines acting in the physical world, is scheduled to begin next year.

The narrative juxtaposes extreme corporate investment in AI development with a call for international governance over existential risk. A central tension exists between the private sector's drive for technological capability and a proposed public/international framework for safety. Son’s focus on the danger of unchecked superintelligence, framed by his enormous financial commitment to the technology, suggests an internal conflict where maximizing technological output seems to clash with ensuring global stability. The simultaneous endorsement of a vision for accelerated scientific research by multiple nations introduces a different vector: the pursuit of advanced capability divorced from immediate risk mitigation strategies. The fact that sovereign states endorse a concept of superintelligence occurring across robotic laboratories highlights a shared, almost utopian vision, which contrasts sharply with Son's fear-based warnings about misuse. This dynamic raises questions about whether coordinated scientific advancement can inherently manage the control mechanisms necessary for existential safety, or if governance structures remain fundamentally inadequate against exponential technological growth. The pattern suggests a split focus between immediate corporate valuation and long-term geopolitical responsibility.
BRIDGE QUESTIONS: If international cooperation is essential to controlling superintelligence, what specific, enforceable mechanisms are missing that could bridge the gap between voluntary declarations (like the Kyoto Vision) and tangible control protocols? How does the current structure of national AI regulations effectively integrate the interests of entities like SoftBank or the drive for pure scientific discovery? What alternative philosophical frameworks for managing exponential power exist that do not rely solely on state-to-state negotiation?