>I could imagine that OAI could be leveraged for single features

That's a possible application which won't be easily copied by the large company. But the value of such features is quite limited.

> I could imagine content moderation, or reading sentiment from user tweets about a company.

I am pretty sure that these services will be provided soon by large companies and OpenAI. They are generic enough to have wide market, and many of these companies already have expertise in them.

>So, no content generation, as that would be unethical without disclosure of AI usage.

Why unethical? I think these systems could create real content which is useful to everyone as well as spam content could be generated by humans.