How OpenAI handles valuable data shared with a GPT I'm developing?

I’m excited about GPTs and have some ideas on how to leverage this to create a useful assistant. First for myself personally and then later perhaps for others.

In the dev day keynote speech Sam Altman stated that “We do not train on your data from the API, or ChatGPT Enterprise, ever.”

I know that I sound a little paranoid here, but does this statement also apply to GPTs built using ChatGPT (i.e. not an assistant created over API)? I’m using my personal account so in this case ChatGPT Enterprise does not apply.

I would like to pass some valuable information to the GPT I’m building but I’m being aware that it is valuable and would not want to take the risk of giving it up for free to OpenAI for them to train future models on.

If you are using custom GPTs via ChatGPT, then yes, assume OpenAI is collecting data on it. As you mentioned, the only instances where they are not doing that is through those with a ChatGPT Enterprise subscription and direct API calls.

Thanks a lot @Macha for clarifying this!

I suppose I’ll go with the Assistant API approach then!

1 Like