DeepMind's Matt Botvinick: AI safety must move from power concentration to checks and balances

schwarzjn_ · x · 2026-09-07

Matt Botvinick, VP of Research at Google DeepMind, argues that while "concentration of power" concerns have rightly entered the AI safety agenda, it's time to go deeper: how to assure a world of separated powers in balanced equilibrium. He offers two models — Acemoglu/Robinson's balance between state and societal power, and the Madison/Montesquieu separation of powers — and frames the key challenge as situating AI labs' power within the Acemoglu/Robinson model both descriptively and normatively, starting from Philip Pettit's notion of non-domination.

Original post →

More from AGI Musings

AGI Musings channel →