i’d like to use GPTs based on some proprietary knowledge we have as well as instructions for assistant behavior we have developped on the API.
i saw on the chatGPT GPTs builder interface an option :
“Use conversation data in your GPT to improve our models” which can be toggled off. This is great!
however does it also apply to the instructions and the file uploaded to configure the GPTs. i have searched and have found no information here.
so my base assumption is that they are NOT kept out of training data… but maybe (and hopefully) am wrong?