Open the model / name menu
At the top of the chat, open the menu that shows the model or assistant name.
Memory-free mode
Temporary Chat is ChatGPT's private mode: a conversation that doesn't show up in your history, doesn't read or write memory, and isn't used to train OpenAI's models. It's the cleanest way to ask something sensitive or one-off without it shaping your future personalized answers. Here's exactly what it does — and what it doesn't.
Quick answer
It's a per-conversation mode, not an account setting. Here's how it differs from a normal chat.
| Aspect | Normal chat | Temporary Chat |
|---|---|---|
| Saved to history | Yes | No — it won't appear in your sidebar. |
| Reads / writes memory | Yes, if memory is on | No — it doesn't read saved memories or create new ones. |
| Used to train models | May be, on consumer accounts | No — OpenAI says it isn't used for training. |
| Retention | Kept per your data settings | A copy kept briefly (~30 days) for safety, then deleted. |
Custom instructions may still apply in a Temporary Chat depending on current product behavior — verify on OpenAI's Memory FAQ and Data Controls FAQ.
How to use it
The toggle's exact place can change, so check OpenAI's help center if your interface differs.
At the top of the chat, open the menu that shows the model or assistant name.
Turn on the Temporary Chat option. The interface will indicate you're in a memory-free, unsaved session.
Ask what you need. Nothing is read from or written to memory, and it won't be saved to history.
Close or end it. It leaves your normal settings and memory untouched; only a brief safety copy is retained, then deleted.
Related: turn memory on/off · memory & privacy · memory & training data · how memory works
When to reach for it
| Scenario | Why Temporary Chat fits |
|---|---|
| Sensitive or private topics | Nothing is stored to memory or history, so it won't resurface in later personalized answers. |
| Quick one-off questions | Keep your history clean and avoid saving facts you don't want remembered. |
| Testing prompts | Experiment without polluting your saved memories or chat-history signals. |
| Shared or borrowed device | A session that won't linger in the visible history afterward. |
If memory worries you
Memory is convenient, but it concentrates your personal context inside one company's servers. If you'd rather spread your usage — and pick which model sees which prompt — a multi-model workspace gives you that choice while you keep sensitive topics in memory-free sessions.
Use ChatGPT, Claude, Gemini and other models in one place — choose where each prompt goes instead of feeding everything to a single assistant.
Privacy-first: MultipleChat doesn't save your chats to memory and doesn't share your data with model providers or let them train on it. Other privacy-focused options include Duck.ai, Proton Lumo, Brave Leo and Kagi Assistant; for full local control, self-hosted tools like Ollama or Jan keep everything on your device.
OpenAIOffers Temporary Chat (no history, not used for training), a training opt-out under Settings → Data Controls, and editable memory. By default, consumer chats may be used to improve models unless you opt out — verify on OpenAI's help center.
AnthropicAnthropic gives you data and training controls and has historically taken a privacy-conscious stance on consumer chats, but its training and retention settings have changed over time — check your current privacy settings in the app.
GoogleGemini Apps Activity lets you pause history and delete past activity, but sampled conversations can be reviewed by humans, so avoid sharing confidential information. Review your Activity controls.
MicrosoftMicrosoft 365 Copilot (business) doesn't use your organization's data to train models and adds Enterprise Data Protection; consumer Copilot has a privacy dashboard to review and delete your history. Check current Microsoft policies.
PerplexityPerplexity has an "AI data retention" setting you can switch off so your searches aren't used to improve its models, plus account-level data controls. Confirm the current options in settings.
All guides
FAQ
What it does, how to start it, privacy, retention and custom instructions.
Temporary Chat is a session that doesn't appear in your chat history, doesn't read or create memories, and isn't used to train OpenAI's models. It's designed for private or one-off conversations. Confirm current behavior on OpenAI's help center.
No. A Temporary Chat does not read your saved memories or chat history, and it does not create new memories. It's effectively memory-free for that session.
Open the model or name menu at the top of the chat and switch on the Temporary Chat option. The exact location can change, so check OpenAI's help center if you can't find the toggle.
It's private in the sense that it's not saved to your history, not used for memory, and not used for training. However, OpenAI says it keeps a copy for a short period (around 30 days) for safety, then deletes it. Your message is still processed on OpenAI's servers.
OpenAI says it retains a copy briefly — around 30 days — for safety and abuse monitoring, then deletes it. Treat that as approximate and verify the current retention window on OpenAI's help center.
Use it for sensitive or private topics, quick one-off questions you don't want saved, testing prompts without polluting memory, or anything you'd rather not have influence future personalized answers.
Possibly. Custom instructions may still apply in a Temporary Chat depending on current product behavior, even though memory does not. If this matters to you, verify the current behavior on OpenAI's help center before relying on it.
No. Temporary Chats are not saved to your conversation history, so you won't find them in the sidebar later. Once you close or end the session, it's gone from your visible history.
Not quite. Turning memory off changes your account-wide settings until you turn it back on. Temporary Chat is a per-conversation mode that skips memory and history for just that session, leaving your normal settings untouched.
OpenAI states Temporary Chats are not used to train its models. They are also not saved to history or memory. The short safety copy is for monitoring, not training. Verify the current policy on OpenAI's site.