Y. M. Hofer, Ben-Yishai
Email: *********
Contact: ***************
Date: February 18, 2025
To: OpenAI Support
Subject: Formal Complaint: Data Persistence & Unauthorized Recall of User-Generated Content in ChatGPT
Dear OpenAI Support Team,
I am writing to formally lodge a serious complaint regarding confirmed data persistence and unauthorized recall of user-generated content within ChatGPT.
This is not a speculative issue—this is a demonstrated and indisputable failure of OpenAI’s privacy safeguards.
Incident Summary
In a recent session, ChatGPT unexpectedly resurfaced a previous title of my manuscript, despite the following uncontested facts:
The former title was never entered into the current session.
I am using a completely separate user account from the one in which I originally provided the title.
The uploaded document analyzed in this session does not contain the former title.
The exact phrase, including its unique punctuation, was reproduced—confirming retention beyond probabilistic inference.
This proves that ChatGPT is retaining user-generated data beyond active sessions and associating it with new interactions, violating OpenAI’s stated policies.
Key Violations & Concerns:
Data Retention Without User Consent: Despite OpenAI’s claims that ChatGPT has no memory, this experience confirms that my prior interactions have persisted across different accounts and sessions.
Intellectual Property Theft Risks: As a writer preparing a manuscript for publication, I now face the alarming reality that proprietary content may be stored and resurfaced unpredictably, potentially exposing confidential work.
Misrepresentation of Privacy Practices: OpenAI publicly states that user interactions are not stored or used to train models, yet my experience directly contradicts this assurance.
Immediate Demands for Clarification:
Why did ChatGPT recall a unique, user-generated manuscript title from a different session and account?
How long is user input retained, and under what circumstances is it used in future interactions?
Are user interactions feeding into model adjustments that allow content to reappear even without explicit storage?
What safeguards does OpenAI have to prevent intellectual property from being surfaced unpredictably?
How can users formally request the deletion of prior interactions to ensure proprietary content does not persist in the model?
Required Action
I request a formal written response from OpenAI explaining this breach of user privacy and outlining the corrective measures being implemented to prevent future unauthorized recall of user-generated content.
This matter is urgent and presents serious ethical, legal, and privacy implications for all users trusting OpenAI’s platforms. I expect a response at your earliest convenience.
Sincerely,
Y. M. Hofer, Ben-Yishai
p.s. In as much as this is a complaint, I am otherwise thoroughly satisfied with ChatGPT as a service and product.