ChatGPT memory can leak across separate projects
ChatGPT did not reliably keep conversations separated by project in a experience. Praana was placed in its own project, yet ChatGPT repeatedly mentioned it in unrelated chats and tried to connect other questions to it.
Claude behaved more consistently: it used the Praana information inside that project but did not bring it into conversations elsewhere. The experience points to a need for better project separation and memory controls, not just stronger AI .
Key points
- Praana had a separate project, but its details appeared in unrelated chats.
- ChatGPT tried to connect unrelated questions back to Praana.
- Claude used the Praana information only inside the project.
- People managing several should check answers for mixed-in details.