OpenAI CEO Sam Altman admits "there's no proven playbook" after former alignment lead suggests the company prioritizes "shiny products" like GPT-4o over safety