ChatGPT: Privacy, Model Training, and more
ChatGPT's privacy policies tend to change constantly. Below are some of the guidelines that are recommended to follow. Please note, while we attempt to provide guidelines on the latest updates, the official site for the current policies and state of ChatGPT is here: OpenAI Help Center.
Model Training
ChatGPT now allows all users to choose if they want to allow their chats to be used to train OpenAI's models. Allowing this means that OpenAI can read the conversations that users have with ChatGPT and learn from those interactions to fine-tune their models. Allowing this also means granting OpenAI access to any data that is shared with ChatGPT, regardless of sensitivity or confidentiality.
Disabling model training will not change the user's experience interacting with ChatGPT. Chats history will still be available.
For logged-in users:
To disable model training, navigate to your profile icon on the top left of the page and select Settings > Data Controls, and disable “Improve the model for everyone." While this is disabled, new conversations won’t be used to train models.
For logged-out users:
To disable model training, navigate to the ? icon on the bottom-right of the page and select Settings > Data Controls and disable "Improve the model for everyone." While history is disabled, new conversations won’t be used to train and improve models.
Temporary Chats
Temporary chats is another way to restrict model training. When enabled, not only will the chats not be used to train models, but they will also not show up in the chat history. OpenAI retains temporary chats for up to 30 days.
Disabling model training is applied to all devices logged in with the user's account.
Learn more about disabling model training and temporary chats.
Learn more about how OpenAI uses data to train models.