SoftBank’s Son warns superintelligent AI could be ‘extremely dangerous’

Masayoshi Son pivoting from pure AI evangelism to public warnings about superintelligence risk is a notable sentiment shift from one of the sector’s most aggressive capital allocators. SoftBank’s Vision Funds have poured tens of billions into AI startups and infrastructure; when Son starts talking about “extremely dangerous” misuse, he is effectively conceding that capability curves are outpacing existing governance structures. Coming from a Japanese business icon on stage with a senior US science adviser, the message will reverberate in both boardrooms and ministries.

Strategically, this puts SoftBank closer to the “accelerate but contain” camp: continue building ever larger AI factories, but recognise that coordination on guardrails is now existential, not optional CSR. That stance matters as Japan positions itself between US and Chinese AI ecosystems, investing in domestic capabilities while relying heavily on US‑designed frontier models and chips. If SoftBank begins to demand stronger safety proofs from portfolio companies or favours investments with clearer governance, it could nudge more capital toward ‘safer by design’ approaches.

For the race to AGI, Son’s warning is unlikely to slow raw capability scaling by itself. But it adds high‑profile business backing to voices arguing that international AI safety regimes should be treated more like nuclear or biotech controls than generic tech regulation.