An OpenAI researcher broke ranks to say that pacing the frontier, as Altman and Amodei suggest, won't be enough

Daniel Selsam, a senior OpenAI researcher, warns that merely slowing AI development ('pacing the frontier') is insufficient to mitigate long-term risks from increasingly situationally aware AI models. He urges more engineering-focused approaches to AI safety, expressing concern about future models that may appear aligned but are not. OpenAI CEO Sam Altman supports pacing but clarifies it means slowing, not stopping, AI progress.