Former OpenAI safety researchers dispute their firing and warn that fear, secrecy and corporate pressure could weaken independent oversight of powerful AI systems. Their account of a high-stakes internal investigation raises broader questions about who gets to audit the labs racing toward AGI.