GPTs uploaded data - level of exposure / privacy

I am trying to understand the level of exposure of additional data that uploaded to a GPTs that I am creating.
If I create a GPTs and upload Excel sheets and additional information that helps GPTs application to work and “understand”, does OpenAI are using this information as part of their training? or this will remain as my “private” application including the additional information (or proprietary information) that I give it?

4 Likes

Great question @tzahif1

  1. In the GPT Editor, in the Configure tab, there is an Additional Settings dropdown at the bottom. Uncheck “Use conversation data in your GPT to improve our models”. This will ensure that submitted messages are not used for training by OpenAI

  2. Adding files in Knowledge does not guarantee privacy. There are a number of other threads on the forum where folks have shared that they are able to download the files in their entirety or pieces of the files. Indeed there is a warning message now that says “Conversations with your GPT may include file contents. Files can be downloaded when code interpreter is enabled.” My general advice is for the time being, if you are going to share your GPT publicly, only upload knowledge that you are comfortable being public.

  3. This will certainly evolve. It’s early and a lot of the general Privacy and Security processes with generative AI are being worked out. GPT Enterprise I believe has much more robust constraints around that but I assume that you are ChatGPT Plus?

There is some additional information on this topic here.

2 Likes

In addition to the very good reply from @cass here is the general information regarding training new models with your data from the help center:

Does OpenAI use my content to improve model performance?

We may use content submitted to ChatGPT, DALL·E, and our other services for individuals to improve model performance. For example, depending on a user’s settings, we may use the user’s prompts, the model’s responses, and other content such as images and files to improve model performance.

In short: if it has to do with the web interface of chat.openai.com treat everything as public.

Here is the link: Data usage for consumer services FAQ | OpenAI Help Center

I understand that ChatGPT conversations’ materials can be used by OpenAI, but I am referring specific to GPTs application that is not public. In this scenario, if I upload files that “helps” my application to work properly, are these materials also consider as public?

Public in the sense that OpenAI may look into them to improve their models unless you specifically opted out.
But I would not be able to look into your files as long as I don’t have the link to the GPT.

So if for example I have some proprietary Excel sheets that helps me improve my exchange trade methodology, I should not use it in any GPTs application if I do not want OpenAI to “learn” from me. Even if my GPTs is private.
Is that correct?

You should use the API instead.
There you have repeated confirmation that the data will not be looked at.

Using API - you refer to GPTs but with API (my python app) instead of creating one using the OpenAI UI ?

Yes. When you build the solution yourself using the API and not the web interface the privacy and security of your data are at a whole other level. Especially with regards to third parties not accessing them and OpenAI not using your data for training.

The developer community here is a great place to find support should your run into any challenges.

Good luck with your project!

plattform.openai.com

1 Like

Thank you for your help and support :slight_smile:

2 Likes

Hello @cass,

In the GPT Editor, I do not have the Additional Settings dropdown at the bottom of the screen in the Configure tab. What I can see at the bottom is attached.
I am a GPT-4 Plus user.

Can anyone help me understand why the Additional Settings dropdown is not there?

image