Joshua Saxe: AI policy progress needs empirical capacity-building, not endless debate

joshua_saxe · x · 2026-11-19

AI safety researcher Joshua Saxe argues in a long post that contested AI policy positions are often well-reasoned on multiple sides and underdetermined by data — so persuasion can't settle them.

He argues the real aim of AI policy should be institutional capacity: gathering the right signals, communicating them to decision-makers, and empirically testing which interventions work. He cites METR, Epoch AI, and cyber threat-intel orgs as models, and calls flexible, empirically grounded institutions an underrated escape hatch from policy-as-endless-seminar.

Original post →

More from AGI Musings

AGI Musings channel →