You can gauge AI safety sincerity by how hard labs police their insiders

yeastsplainer · x · 2026-09-13

The author proposes a test for AI safety sincerity: a rogue insider researcher poses vastly greater danger than any retail user, so the severity of surveillance and policing applied to insiders reveals how real a lab's safety concerns actually are.

Original post →

More from AGI Musings

AGI Musings channel →