Ex-Policy Head Warns: AI Companies Normalizing Automated R&D Risks

Miles_Brundage · x · 2026-08-10

Former OpenAI policy researcher Miles Brundage warned that AI companies are "getting away with" risky automated AI R&D and recursive self-improvement (RSI) by normalizing it gradually.

Citing the sociological concept of the "normalization of deviance," he argued that the industry's gradual shift—from IDE autocomplete to coding agents with tool use, and eventually to automated AI R&D—desensitizes stakeholders to catastrophic risks, making an obviously insane idea seem acceptable step by step.

Related event: Ex-OpenAI Adviser Brundage Warns of AI Regulatory and Safety Blind Spots(15 posts)→

Original post →

More from AGI Musings

AGI Musings channel →