Clarifying the AI safety split: safety testing time vs long internal deployment

JacquesThibs · x · 2026-09-15

Jacques Thibodeau clarifies his earlier thread: some EA/AI safety folks have been confused about external releases partly because they objected to safety testing being shorter — not to the longer internal deployment phase after testing, which he sees as the main risk source.

Related event: AI Safety Circles Debate Whether Early Model Release Is Safer(2 posts)→

Original post →

More from AGI Musings

AGI Musings channel →