Is OpenAI's adult content policy actually contradictory?

TL;DR: OpenAI’s Usage Policies don’t seem to ban consensual adult sexual content, but their account-ban page lists “nudity and sexual content” as a common violation. I’m trying to understand whether Astra refusing adult-related tasks is just model behavior, or something that can actually put the account at risk. Either way, I don't understand why 5.6 Sol never ever refused to do anything, while Astra just goes "Nope" by only reading a single skill, it makes no sense. Hi everyone, as a non-native English speaker, and ADHD Ultra guy, I resumed my researches with GPT and tried to make it as least sloppy as possible. I'm confused about OpenAI's policy on adult content. The current Usage Policies, effective October 29, 2025, ban non-consensual sexual content, minors, sexual violence, etc. But they don't seem to ban consensual adult sexual content in general. At the same time, OpenAI's Help Center article about account bans, updated much more recently, lists: as a common policy violation. So which one is actually enforced? I work in the adult creator industry. GPT-5.6 Sol handles a lot of my workflows fine, but Astra refuses pretty basic stuff once adult content is involved. For example: filling a Google Sheet from existing content descriptions copying an existing script from a Sheet into another tool No new sexual content even needs to be generated. I was super hyped by everything said about computer used being improved, but well... can't use it. But most importantly, I care about is whether triggering these refusals can put my account at risk. I also don't want to keep testing different setups if that could be interpreted as bypassing safeguards. Is there any clear OpenAI documentation explaining the difference between: “this model won't do this” “asking for this is actually a policy violation” And why would OpenAI unallow the model to do a task that is not forbidden by their own conditions ? Because right now the docs seem to say two different things.