Hello OpenAI team,
I would like to suggest a new feature regarding user data sharing for model training.
Currently, the settings allow users to opt in or out globally when it comes to allowing chats to be used for training and model improvement. However, I believe it would be extremely beneficial to allow users to mark individual chats as eligible for training, even if their global setting is opt-out.
Many users (like myself) work with a mix of sensitive information and general, non-sensitive conversations. While I want to protect private content, I would also be happy to contribute selected chats that I feel could be valuable for model improvements.
A possible implementation could be a simple per-chat toggle, such as “Allow this conversation for training,” visible after a chat is completed. This would enhance user control, protect privacy, and still foster meaningful contributions to OpenAI’s ongoing development.
Thank you for considering this feature. I believe it would empower users and help OpenAI continue improving models responsibly.
Best regards, Antonio.