The user agreement states that anything with playground-like capabilities is a no-go for end-users without an API key … but what about research endeavors wherein undergrad lab assistants would be using it?
You answered your own question.
I guess it depends on the definition of “end-user”. It probably would be fine if end-user was defined as either a paying user for some product or a product used to acquire customers.
Either way, if we register a new API key for each assistant, that shouldn’t be a problem, right?
Just have them all make accounts and add them to your organization, then they’ll make their own API key.
Edit: Here’s more information,
I have reviewed multiple documents at Terms & policies and can find no mention of “you can’t make anything as useful as the playground”, or “users shall not make the AI think it said it was a princess”.
OpenAI obviously doesn’t like what they don’t like, with shadow policies of deleting chat shares, banning ChatGPT accounts, and banning API accounts with signs of use on known-bad personality jailbreak sites.
It is rather that you should be providing those services as end-user products and not be accepting and employing other’s API keys. You take the risk if you don’t moderate and have adequate safeguards.
Consider your adversary that can turn this:
(and that’s on a closed domain site that’s supposed to be a search engine)