Summary
I would like to request a per-Project privacy setting that allows users to exclude specific ChatGPT Projects from model training and improvement, while keeping the global data-sharing setting enabled.
The use case
I am generally comfortable contributing my conversations to model improvement. However, not all conversations have the same level of sensitivity or purpose.
For example, I may want my everyday conversations, technical discussions and general questions to remain eligible for model improvement, while keeping a specific creative-writing Project entirely excluded.
Today, the global setting does not offer this level of control.
Proposed functionality
Add a setting at the Project level, such as:
-
Follow global setting (default)
-
Exclude this Project from model training and improvement
The global account setting would remain unchanged.
If global data sharing is enabled, Projects explicitly excluded by the user should remain excluded.
The Project-level restriction should take precedence over the global setting.
Ideally, this exclusion should cover all content associated with the Project, including its conversations and uploaded files.
Why this matters
Users should not have to choose between contributing to model improvement in general and protecting specific areas of their activity.
A per-Project opt-out would provide more granular control, respect different privacy expectations and encourage users to keep global data sharing enabled when they are comfortable doing so.
This is complementary to the existing request for granular data-sharing controls:
That request already discusses selective sharing at conversation, Project and message level. My proposal emphasizes the complementary use case: global opt-in with explicit per-Project opt-out.
Thank you for considering this.