DKokotajlo says they plan to attend a November 19, 2026, rally outside Anthropic, backing a campaign to bar AI models stronger than today’s from building their successors. Organizers list the Ban RSI protest for 11 a.m. to noon PT at 500 Howard St. in San Francisco. Source Source
The target is recursive self-improvement, or RSI, which the campaign defines as AI doing the work of building more powerful AI. Organizers warn that repeating that process could accelerate development beyond humanity’s ability to understand and manage the risks. Source
“Our line is today’s level. Nothing stronger may be put to work building its successor,” the campaign’s FAQ states. A brake on the next round of acceleration, rather than an order to unplug every current research tool. Source
In their November 18 post, DKokotajlo called recursive self-improvement “a stupidly dangerous idea” and rejected the competitive rationale that someone else would pursue it anyway. They wrote: Source
Anthropic (like OpenAI before them) styles themselves the good guys in the race, but IMO there are no good guys in this race. Source
The campaign wants lawmakers to enact a ban. It also asks Anthropic to announce a plan to refrain from RSI and urge other AI companies to follow suit or lobby for legislation. Source
The campaign’s FAQ does not spell out how “today’s level” would be measured or enforced. The slogan is considerably shorter than the rulebook would need to be. Source
Comments