Last verified

Are ChatGPT chats private?

Not really.
Stored by default · trains models unless you opt out

Not by default. On a personal plan your conversations are stored in your account, and may be used to improve OpenAI's models until you turn that off. They are not sold or shown to advertisers — OpenAI states that plainly. Three controls change the picture: the training toggle, Temporary Chat, and deletion. Business, Enterprise and Edu workspaces start with training off.

Stored by default — chats stay in your account until you delete them Used for training by default on personal plans, until you switch it off Not sold to advertisers — no chats, memories, name, email, IP or precise location Temporary Chat is excluded from training and deleted after 30 days Business / Enterprise / Edu are not used for training by default
§ 01 / SOURCE

Where this answer comes from.

Methodology →
PER VENDOR DOCS · AUG 21, 2026

"Private" is two questions, and OpenAI answers them differently.

Private from OpenAI? No, not by default. The Data Controls FAQ describes a setting called "Improve the model for everyone" that is on unless you turn it off. With it on, your conversations may be used to train models. Turning it off is available signed in or signed out, applies to the entire account across devices, can be changed at any time with no restrictions, and leaves your chats visible in your own history. It is not retroactive.

Private from advertisers? Yes, and unusually specifically. Since OpenAI began selling ads it has published exactly what advertisers receive: aggregated views and clicks. The ads FAQ states that advertisers never receive your chats, chat history, memories, name, email, precise location, IP address or sensitive information, that conversations are not shared with advertisers, and that data is never sold to them. Ads are also excluded from health, mental-health and political conversations, and from Temporary Chats.

Temporary Chat is the strongest consumer control. Per the same FAQ, Temporary Chats are not used to train models, do not appear in history, create no memories, are deleted from OpenAI's systems after 30 days, and may be reviewed only to monitor for abuse. Files uploaded inside one never reach your Library. That last clause — reviewed for abuse — is the honest limit of the word "private" here, and OpenAI publishes it rather than hiding it.

One caveat OpenAI flags itself and we would rather repeat than bury: turning off memory and personalization does not disable safety features, which may use limited, safety-relevant context in rare high-risk situations. If your definition of private is "no system ever reads this", no consumer chat assistant meets it — theirs included.

Stored by default: YesTrains models by default: Personal plans: yesOpt-out available: Any timeOpt-out retroactive: NoSold to advertisers: NoTemporary Chat retention: 30 days
!
Per vendor docs — we did not run this test. The answer above rests on the source below, read on . Where we run a capability ourselves, this section carries the terminal transcript instead.
Source OpenAI Help Center — Data Controls FAQ (updated 3 days before we read it) https://help.openai.com/en/articles/7730893-data-controls-faq
§ 02 / WHY

The three switches, in order of strength.

2.1
Temporary Chat — strongest, and per-conversation

If a conversation should not exist afterwards, start it as a Temporary Chat rather than deleting it later. It is not saved to history, creates no memories, is not used for training, and leaves OpenAI's systems after 30 days.

The trade is that you lose the thing chat history is for: continuity. Nothing carries over, and you cannot come back to it tomorrow. It is the right default for anything involving somebody else's personal data, an employer's confidential material, or health and finance details you would not paste into a shared document.

2.2
The training toggle — account-wide, not retroactive

Settings → Data Controls → turn off Improve the model for everyone. Your conversations remain in your history and remain usable; they stop feeding model training from that point on.

Two details matter. It is account-wide, so you do not have to repeat it on mobile. And it is not retroactive — conversations you had before flipping it were already eligible. If you previously opted out by writing to support or via the privacy form, OpenAI states that request is still honoured on your account.

2.3
Deletion — cleans the record, slowly

Deleting a chat removes it from your view immediately and from OpenAI's systems within 30 days, with the usual carve-outs for de-identified data and legal retention. Archiving does none of this — archived chats are simply hidden. The mechanics are on our delete-history page.

Memory is a fourth, separate store: what ChatGPT remembers about you across conversations lives in Settings → Personalization and survives a history purge until you clear it there.

2.4
The plan you are on changes the default

The answer to this question is genuinely different depending on who pays. On personal plans — Free, Go, Plus, Pro — training is on until you turn it off. On Business, Enterprise, Edu and the API platform, OpenAI states it does not use your content to train models by default, and those plans carry additional data controls.

That is worth knowing before you paste work material into a personal account: the same text has a different default depending on which login you used. If your organisation has a workspace, use it. If you are choosing between tiers for this reason specifically, Plus vs Team is the comparison that covers it.

Frequently asked.

Quick follow-ups people search after this question.

Q · 01 Does OpenAI sell my ChatGPT conversations? +
No. OpenAI's ads FAQ states that conversations are not shared with advertisers and that user data is never sold to them. Advertisers receive aggregated performance figures such as total views and clicks, and no chats, memories or personal identifiers.
Q · 02 Can OpenAI employees read my chats? +
OpenAI does not describe unrestricted access, but it does document that Temporary Chats "may be reviewed only to monitor for abuse", and that safety systems may use limited safety-relevant context in rare high-risk situations. Treat the chat box as a hosted service, not as an encrypted notebook.
Q · 03 How do I stop ChatGPT training on my conversations? +
Settings → Data Controls → turn off "Improve the model for everyone". It applies to your whole account across web and mobile, is changeable at any time, and leaves your chats in your history. It is not retroactive.
Q · 04 Are ChatGPT chats confidential enough for work material? +
On a personal plan, that is your employer's decision rather than a technical one — training is on by default and the content is stored. Business, Enterprise and Edu workspaces are not used for training by default, which is the configuration most organisations expect.
Q · 05 Does Temporary Chat leave any trace? +
It is not saved to your history, creates no memories, is not used for training, and is deleted from OpenAI's systems after 30 days. It may be reviewed for abuse monitoring within that window. Files uploaded in a Temporary Chat are not saved to your Library.
Q · 06 Does deleting my chats undo the training? +
No. Deletion removes the conversation from your account and from OpenAI's systems within 30 days, but training that already happened cannot be reversed by deleting the source text. The training toggle only affects what happens from the moment you flip it.