Frontier AI should be credibly destroyable or publishable by n-of-k actors

DavidSKrueger · x · 2026-11-17

David Krueger argues frontier AI should be developed so it can be credibly destroyed or published with the agreement of n of k actors. Today there is no way to credibly destroy a model even if it's found to be too dangerous during or after development — a governance and safety gap he calls unacceptable.

Original post →

More from AGI Musings

AGI Musings channel →