In the bug thread, @objmihail2 cited OpenAI’s help center:
Once you’ve enabled the chat history, you can still opt out of our use of your data to improve our services by filling out this form: Opt-Out Form . Upon submission, new conversations will not be used to train our models.
When you turn chat history off via the switch, this is visually confirmed to you. But, unfortunately, when you use the opt-out form, you get no visual confirmation of that in the ChatGPT interface. I guess I’m not the only one who is made to question the effectiveness of the “setting” because of that.
So, getting a visual confirmation of having submitted the opt-out form in the ChatGPT interface would be a sensible alternative to the current buggy behavior with the switch to turn off training and chat history. Of course, the switch could also just do the equivalent of filling out the form, but this may be too much of an incentive for users from OpenAI’s perspective.