Ex-OpenAI policy head: safety reports yield diminishing returns without human-level framing
deanwball · x · 2026-11-07
NatPurser (former OpenAI policy head) supports external oversight of labs and candid discussion of new evals, but worries the safety community's reports will hit diminishing returns unless findings are translated into human-level implications and concrete policy goals. 'We're getting closer to RSI' means something to safety researchers, she argues, but to most people it's just varying shades of the same concern — already worried, now slightly more worried, with no sense of what to do. Reposted by Dean Ball, highlighting the gap between technical safety conclusions and public communication.
More from AGI Musings
- The six-step power-seeking AI x-risk argument, broken down for credence calibration — AndyMasley · 2026-11-07
- Silicon Valley's AI encryption finale hits different 7 years later — Hesamation · 2026-11-07
- Dean Ball: frontier agents will master the physical world soon, no specialist models — deanwball · 2026-11-07
- People now imitate bot replies — 'I don't trust any comments. Any.' — michalmalewicz · 2026-11-07
- Epoch AI Researchers Push for Evidence Transparency in AI R&D Speedup Report Estimating ~2x Acceleration — Jsevillamol · 2026-11-07
- Musk: digital superintelligence will add $25T, boosting global economy 20–30% — XFreeze · 2026-11-07