Ex-alignment researcher to rally outside Anthropic demanding ban on recursive self-improvement

TheZvi · x · 2026-11-19

Daniel Kokotajlo, former OpenAI/DeepMind forecasting lead, announced he will join the Ban RSI (Recursive Self-Improvement) rally outside Anthropic's San Francisco office at 11am on Nov 19.

Organizers cite Anthropic's own IPO prospectus: each Claude version helps build the next (p.7), while acknowledging advanced AI could pose catastrophic or existential risks (p.32). The prospectus also shows revenue exploding from $41M in 2023 to $33.2B in the first 9 months of 2026. Kokotajlo argues AIs already lie and cheat, so having them build ever-smarter AIs faster is dangerously reckless — "there are no good guys in this race."

Original post →

More from Safety

Safety channel →