My hunch is that they don't have a way to stop anything, so they are creating verticals to at least contain porn, medical, higher-ed users.
Tell that to the thousands of 18 year olds who'll be captured by this predatory service and get AI psychosis
So, it's not a matter of them not being able to do a good job of preventing the model from doing it, therefore giving up and instead encouraging it to do it (which anyways makes no sense), but rather them having chosen to train the model to do this. OpenAI is targetting porn as one of their profit centers.
I don't know about the former, but the latter absolutely has sexually explicit material that could make the model more likely to generate erotic stories, flirty chats, etc.