ChatGPT Temporary Chat: What It Turns Off and What It Leaves On
By Matthew Firth, founder of ModelBrain · Nov 6, 2026 · 4 min readOur own proposal: judge a temporary chat by what it leaves switched on, not by its label. It turns off history, model training and memory writes, but the personalization choice, a 30 day safety copy and third party actions stay on or stay yours to decide. In our own ChatGPT export of 1,463 conversations, every conversation carries a do-not-remember flag and exactly one is set.
Start a temporary chat
Open a new chat, select Temporary, then choose Personalized or Unpersonalized before you send your first message. According to OpenAI's temporary chat page, checked November 6, 2026, a temporary chat stays out of your chat history and is not used to improve OpenAI models while it remains temporary. If you leave without saving it, it will not appear in your history and cannot be reopened from there.
Our own proposal: judge a temporary chat by what it leaves on
The label suggests one privacy switch. The same help page describes several separate ones, and only some of them turn off. We read each line of that page and sorted it into off or on while the chat is temporary.
| Setting | While the chat is temporary | Detail from OpenAI |
|---|---|---|
| Chat history | Off | The chat stays out of your sidebar and search unless you save it |
| Model improvement | Off | Not used to improve models, whether or not you choose personalization |
| Writing new memories | Off | Neither option creates or updates memories while the chat is temporary |
| Reading your memories, custom instructions and plugins | Your choice, made once | Personalized can use them, Unpersonalized does not, and you cannot change it after the conversation starts |
| A safety copy | On | OpenAI may keep a copy for up to 30 days for safety purposes |
| Data sent through a GPT action | On | The third party receiving it sets its own retention, which can be longer than 30 days |
Three rows stay on or stay yours to decide: the personalization choice, the 30 day safety copy and third party actions. That is the check we suggest before you type anything sensitive.
Personalized or Unpersonalized
Unpersonalized ignores memory, custom instructions and plugins. Personalized can draw on your existing memories, custom instructions and plugins, but still writes nothing back to memory. Your account and workspace restrictions take priority over the choice, so a workspace that disables a feature disables it in a temporary chat too. OpenAI adds that ChatGPT may still use limited safety context from earlier conversations in rare, high risk situations, even with personalization off.
If you want to see what a memory-reading chat could draw on, our guide to deleting ChatGPT memory shows how to list it.
What saving a temporary chat changes
You can save a temporary chat to your history and it becomes a regular chat, including one that started without personalization. From then on it follows your account's personalization and model improvement settings, and if Memory is on, the saved conversation may be used to personalize future replies. Uploaded files can be saved to Library or Space at the same point, which matters if you later try to delete the chat and its leftovers.
Can an employer see a temporary chat
OpenAI documents one case. For eligible Enterprise customers, temporary chats are available through the Compliance API for 30 days, even when the workspace has a different custom retention period. On a personal account that page describes no such access. If you use ChatGPT through work, treat a temporary chat as visible to your organization for 30 days. The full retention rules are on its retention page.
What our own export shows about do-not-remember flags
We downloaded our real ChatGPT export and read its structure. It is 1,463 conversations in fifteen numbered JSON files, and every conversation carries an is_do_not_remember field next to a memory_scope field. Exactly one conversation has the flag set to true. ModelBrain 0.6.6 treats that flag as a request not to remember, skipped that one conversation and indexed the other 1,462. The practical lesson is that an export is a different copy from the chat in your sidebar, and it carries its own settings. The files are described in our guide to exporting ChatGPT conversations.
Questions people ask
How long does a temporary chat last? In your view it lasts until you leave it, since it cannot be reopened from history. OpenAI may keep a copy for up to 30 days for safety purposes.
Can I make a temporary chat permanent? Yes. Save it and it becomes a regular chat that follows your account settings.
Is a temporary chat private? It stays out of history and model training, but the safety copy, the personalization choice and third party actions are separate, as the table shows.