David Robinson, who led safety transparency efforts at OpenAI, quit because he believes the company’s technology is not safe enough. A Rhodes scholar and former Biden White House advisor, Robinson joined OpenAI thinking AI safety risks were 'nuts' but changed his view after three years. He argues OpenAI lacks the safety culture needed to prevent catastrophic risks, comparing potential harms to a nuclear meltdown but far worse.
He says the industry operates like a startup despite building dangerous systems, with insufficient internal controls and redundancies. Robinson notes public safety reports from OpenAI, Hugging Face, and Anthropic show problems, but external observers may overestimate their safety rigor. He distinguishes between organizational excellence, technical safety feasibility, and the right metaphors for risk—like nuclear power.
Robinson insists AI alignment is not merely an engineering problem. His resignation highlights growing internal dissent over AI safety practices in the industry.
Source: New York Times Top Stories · Summarized by HeadlinesBriefing