I am not really a tech person and don’t fully understand what the term API even means, so apologies if this is obvious to others. But I was looking at Open AI’s data use policy (How your data is used to improve model performance | OpenAI Help Center), which makes a distinction between OpenAI’s API and Open AI’s “non-API consumer services”. The policy makes clear that ChatGPT and DALL-E are both considered non-API consumer services. But the policy doesn’t say how the playground is classified. I am interested in knowing what the data use policy is for information entered into the playground, but this doesn’t seem to be explained anywhere. Can anyone clarify?
Is the playground considered a "non-API consumer service"? Or is it part of the API?
Welcome to the forum.
I believe Playground would fall under API services and not consumer… I could be wrong, tho…
There is no reason to guess, you can simply test it and confirm yourself.
So, just for you @JC1 , I just did this simple test for you by going to the playground and creating a new completion with a model I rarely use,
Then, I waited 5 minutes and looked at my account token usage, and “abracadabra”, there it is!
This clearly shows, without guesswork, that when you use the playground, you are charged just like you are using the API (per usage basis, not retail consumer month subscription)
Hope this helps.
Thank you! So when this page (OpenAI API) says that “Any data sent through the API will be retained for abuse and misuse monitoring purposes for a maximum of 30 days, after which it will be deleted (unless otherwise required by law)”, you’re pretty confident that this applies to any data entered in at the playground?
Yes of course. 100% confidenr.
lol - I only just now noticed this forum post (Getting started - where do I begin exploring the API?) by OpenAI staff, in which they explicitly state that the playground is part of the API.
Still, thank you for your help!
I disagree with @ruby_coder . Nearly all modern web services call apis on the backend. So the fact that the API token use goes up doesn’t really tell me anything.
The documents could be saying, “If you make direct calls to the API (via curl or python for example) then you don’t contribute to improvements in the model”.
But any service like Playground or ChatGPT that makes calls the api for you and has the ability to check your accept/reject interaction could do that.
In fact, there is a button at the bottom of the playground, thumbs up thumbs down. And if you click it, a toast pops up that says “Thanks, your input helps us improve our models”.
If you want more pseudo-proof, ask Playground if it can analyze the ToS and tell you if “any inputs or interactions used in playground are used to improve the model”.
Playground will regurgitate the docs and tell you you are fine.
ChatGPT will intelligently parse and reason and tease out that the Api-data-privacy-policy only applies to the Files endpoint which is used in fine-tuning only. It goes on to say that Playground is NOT covered by the non-API terms and that user feedback is used to improve the model.
Someone at OpenAI needs to weigh in so we have clarity.
Through other integration partners I learned that as of Mar 1, all direct API calls are opt-out transactions concerning use of content for additional training purposes.
In contrast, all OpenAI web apps are opt-in and use your information.