I’m posting here out of urgency and deep concern. Between April 19 and April 28, 2025, I experienced a critical ChatGPT memory and export failure that resulted in the total loss of a professional writing project — specifically, the first half of a novel I had been building intensively with GPT-4o.
Despite multiple attempts to save/export my work — including a session where the assistant stated my book was 90% compiled — the process failed, and my data vanished. I also encountered system-level errors like:
• UnicodeEncodeError: ‘latin-1’ codec can’t encode character
• OSError: encoding error in FPDF output
• TypeError: an integer is required (got type str)
These were not user-side issues. A glitch occurred that wiped memory and blocked me from preserving what had already been generated. I have already submitted full documentation, error logs, and timeline to OpenAI Support.
While I was told the issue would be escalated to technical support, I’ve since received mixed follow-up messages and automated responses, some of which don’t align with that promise. As of now, I have not received any confirmation that my system logs are under review.
This is not an experimental or casual use case — this novel was my professional work in progress. The content was original, emotionally significant, and irreplaceable.
I am asking OpenAI technical support to:
1. Confirm whether this case has been escalated.
2. Clarify if my memory and export logs are being reviewed.
3. Retrieve the stored conversations (within the 30-day window stated in OpenAI’s policy).
4. Provide a point of contact for consistent updates — not a rotation of auto-generated replies.
I’m making this public so that others facing similar issues feel less alone and so the matter receives real attention. I hope OpenAI takes its responsibility seriously when memory and system reliability are promoted features.
Thank you.
Jennifer Saba