sull Warns Against Restricting Open-Weight Models: Responsibility Isn't Global Advantage

sull · x · 2026-09-14

In an X thread, sull pushes back on AI slowdown framing: the point isn't "self-improvement" but ensuring humans can understand what's being built and stay in control. Human-led AI advancement is plenty for now, businesses would be fine with a frontier-lab slowdown. His sharpest warning: don't restrict open-weight model access — conflating responsibility with global competitive advantage "would be the tell."

Related event: Ex-OpenAI safety researcher: slowing AI is about control, not self-improvement(2 posts)→

Original post →

More from AGI Musings

AGI Musings channel →