Zvi Mowshowitz argues that safety culture at frontier AI labs is broken

Zvi Mowshowitz argues that frontier AI labs lack a deep and robust safety culture, with OpenAI's problems running deeper than Anthropic's.

He cites David Robinson's resignation and the firing of three OpenAI safety researchers, along with misalignment reports from both companies.

Robinson argues that alignment is a science problem and that labs need safety practices borrowed from fields like aviation and nuclear power.