Privacy and Trust: A Proposal for Truly Confidential Chats
I’d like to raise what I believe is not just a feature request, but an important trust and privacy issue.
Before explaining my proposal, I’d like to ask one simple question.
If someone gained access to your ChatGPT account for just five minutes, could they discover your most private conversations simply by asking the right questions—even after you had deleted those chats?
That question made me perform several tests.
I instructed ChatGPT that an entire conversation should be treated as strictly temporary.
I explicitly asked it to:
never store anything from that chat as memory;
never use it to personalize future responses;
never reference it outside that conversation;
and, if I ever asked about it from another chat, to simply answer that it had no knowledge of it.
I then deleted the conversation.
Later, from a completely different chat, I asked questions related to the deleted conversation.
The result surprised me.
Instead of acting as though that conversation no longer existed outside itself, ChatGPT was still able, in some situations, to answer using information from it.
I’m not questioning how OpenAI stores data internally.
I completely understand there may be security, legal or operational reasons why deleted conversations are retained for some time.
That isn’t my concern.
My concern is something different.
Users currently have no way to guarantee that a conversation will never exist outside the conversation itself.
To me, this is one of the biggest trust gaps in ChatGPT.
Imagine discussing:
your sexuality;
a mental health issue;
confidential work information;
family problems;
legal advice;
or any deeply personal topic.
ChatGPT is at its best when users feel comfortable telling the complete truth.
But complete honesty only exists when there is complete trust.
If users are unsure whether a confidential conversation might later be referenced from another chat, many will inevitably hide information or avoid discussing the subjects where ChatGPT could help them the most.
My proposal
I think ChatGPT should include a Conversation Isolation Mode.
When enabled:
the conversation becomes permanently isolated from every other conversation;
nothing inside it can ever be used as memory;
nothing can influence future conversations;
if another chat asks about it, ChatGPT simply replies that it has no access to that information;
deleting that conversation guarantees it can never affect future responses.
Not because the AI “forgot”.
Because the system guarantees that isolated conversations stay isolated.
Why this matters
Privacy is important.
Memory controls are important.
Deleting chats is important.
But none of those are the same thing as conversation isolation.
I believe users should have the right to decide that certain conversations never exist outside the conversation in which they were created.
That single capability would dramatically increase my trust in ChatGPT, and I believe many other users would feel exactly the same.
I’d genuinely like to know what both the community and the OpenAI team think about this idea.
Thank you for reading.