They're not claiming they don't moderate, though. Where are you getting that? A common complaint about ChatGPT and even their open weights models is that they're too censored.