According to remarks at the Semafor conference, OpenAI CEO Sam Altman addressed recursive self-improvement (RSI)—enabling AI to help develop stronger next-generation models and progressively enhance itself. When asked whether OpenAI has achieved this capability, Altman explained that like AGI (Artificial General Intelligence), RSI has no clear-cut boundary; the company does use its AI to accelerate next-generation model development.
However, he clarified that OpenAI has "absolutely not" implemented unsupervised AI self-design and iterative improvement without human oversight. Altman noted OpenAI could advance RSI further but is deliberately slowing down due to safety concerns, stating the company cannot yet guarantee such approaches are safe.