Manheim: without distinguishing safety from alignment, AI policy is impossible

davidmanheim · x · 2026-09-22

Continuing the terminology debate, the counterpart holds that "alignment" is indispensable in technical discussions, just not in general policy discourse, and that jargon misappropriation is unavoidable—one can only politely correct it. David Manheim strongly disagrees: if we cannot make the substantive point that a system's safety differs from its alignment, we cannot make reasonable policy decisions about future systems—which is exactly the current debate.

Related event: AI Safety Researchers Debate Whether "Alignment" Belongs in Policy Talks(4 posts)→

Original post →

More from AGI Musings

AGI Musings channel →