dots memory: what it stores, shares and cannot delete

A dot’s memory is shared with ChatGPT in both directions and cannot be edited piecemeal: there are no controls to view, delete or modify individual memories. The only thing that deletes a dot’s context is deleting the dot — and even then, some data survives because it lives elsewhere.

What it receives and what it contributes

The exchange is bidirectional and continues after initial setup:

Direction	What happens
ChatGPT → dot	The dot receives memories and recent context from your conversations
dot → ChatGPT	Conversations with your dot can contribute to ChatGPT memory

On top of that, the dot creates its own memories, including ones formed from connected apps. It can do so without you asking: that is part of proactive research.

That last part is the most surprising. A dot connected to your apps can form recollections about information you never explicitly asked it to look at.

The limit worth knowing before you use the product

Here is dots’ least intuitive limitation:

You cannot view, delete or modify individual dot memories. Nor the specific details entering its context from plugins. OpenAI states this explicitly.

So there is no equivalent of “delete this one recollection”. The only levers available are:

• Delete the dot → removes its entire context.
• Turn off memory in ChatGPT → stops the exchange, but does not delete what the dot already received.
• Manage ChatGPT memory → affects what ChatGPT remembers from your other conversations, not the dot’s own context.

If you are evaluating dots for a case where you need granular control over what the system remembers, this is a decision input, not a footnote.

What it does NOT store

There is good news in the design:

• Credentials — not retained in the dot’s context.
• Images and screenshots — neither.
• Passwords — when a dot needs to sign in to a supported site, it pauses and you enter the credentials in a secure form that sends them directly to the browser environment. The model does not see them.

How long it keeps context

For as long as you keep the dot. There is no defined expiry window: context persists until you delete it.

Files the dot creates or uses elsewhere — ChatGPT Library, your computer, a connected storage service — follow those locations’ retention policies, not the dot’s.

What gets deleted and what survives

The table worth having at hand before a Reset:

Item	Deleted with the dot?
The dot and its context	Yes
The dot’s saved memories	Yes
Scheduled tasks	Yes
Files it created	No — stored separately
Codex threads it created	No
ChatGPT conversations it created	No
Memories from your other conversations	No — managed separately

Disconnecting an app deletes nothing

Another surprising behaviour: disconnecting a plugin stops new access, but does not remove information the dot already built into its context. If you disconnected an app to limit what the dot knows, you did not achieve that.

To genuinely delete that information, you have to delete the dot.

How to think about a dot’s privacy

Three practical rules that follow from all of the above:

1. Before connecting an app, assume its information may remain in the dot’s context permanently. Connecting is a near-irreversible decision short of deleting the dot.
2. If granular control over memory is a requirement, dots does not meet it today. There is no way to audit or prune what it remembers.
3. Shared memory with ChatGPT widens the surface. What the dot learns can influence your conversations, and vice versa.

How to request access to or deletion of your data

Independently of the dot, OpenAI provides a privacy portal and an email for requests: privacy.openai.com and [email protected], even without an account. Depending on where you live, your rights may include accessing, correcting, deleting or receiving a copy of your personal data.

Memory is the part of the product with the fewest available controls. If your use case depends on being able to audit what the agent knows, decide that before adopting, not after.