OpenMined's Trask: AI safety debate boils down to who unilaterally controls superintelligence

iamtrask · x · 2026-09-14

Andrew Trask (iamtrask) argues that 99.999% of the AI safety debate just circles two questions: who should be unilaterally controlling superintelligent AI, and how can we ensure AI isn't unilaterally controlling itself.

In a follow-up reply he breaks it into three cases: closed-source AI is unilaterally controlled by a company; open-source AI is unilaterally controlled by anyone with a copy; rogue AI is unilaterally controlled by the AI itself.

Related event: OpenMined founder reframes AI safety debate around unilateral control(7 posts)→

Original post →

More from AGI Musings

AGI Musings channel →