23 'Low-Regret' Recommendations for AI Policy and Pacing

AndyMasley · x · 2026-08-14

Noahpinion highlights a comprehensive guest post by Tim Fist and Saif Khan from the Institute for Progress, outlining 23 specific, 'low-regret' recommendations for AI policy.

Following up on a previous discussion about whether to deliberately 'pace' AI self-improvement, the authors argue that while the severe risks of automated AI R&D are uncertain, preparatory policy action is warranted.

Key proposals include specifying thresholds for when automated R&D poses severe risks. If a threshold is exceeded, policies should incentivize AI companies to reallocate resources away from risky research toward safety and benefit diffusion. The authors warn that without preparation, political backlash could lead to poorly reasoned measures, such as broad bans on new data centers.

Original post →

More from AGI Musings

AGI Musings channel →