Are ChatGPT Chats Private? What AI Memory Stores

AI memory is private only within the limits of the account, settings and provider’s data practices. Turning memory off can reduce personalization, but it does not automatically erase conversations or stop every form of data retention. Training controls are another decision.

If you are asking “is AI memory private,” start by separating four things: what the assistant uses in future replies, which conversations and uploaded files remain in history, whether content may improve models, and how long the provider retains data. This guide covers personal consumer accounts. Workplace and school accounts can have different administrator controls and retention rules.

Is AI memory private when memory is switched off?

Does an AI writing assistant store your documents, and for how long?

Yes. ChatGPT, Gemini and Claude keep a file you upload for as long as the chat, project or activity record that holds it exists. Deleting it is your job, and ChatGPT and Claude then take up to 30 days to clear their copies.

ChatGPT can keep a file in two places: inside the conversation and in your Library. OpenAI says deleting a chat does not delete a file that is still saved in Library, so remove both. A deleted chat is scheduled for permanent deletion within 30 days. Files in a project stay until you delete the file or the project, and knowledge files in a custom GPT stay until you delete the GPT. Some accounts now see Space in place of Library. OpenAI lists the details in Chat and file retention in ChatGPT.

Gemini works on a timer instead. With Keep Activity on, Google saves your chats and what you share with Gemini, including files and photos, in Gemini Apps Activity, and deletes that activity after 18 months by default. You can change it to 3 or 36 months, or keep it indefinitely. With Keep Activity off, chats stay with your account for 72 hours. While Keep Activity is on, a sample of your uploads can also be used to improve Google services, with human reviewers involved. The Gemini Apps Privacy Hub has the full list.

Claude treats an uploaded file as part of what you send in the chat. When you delete a conversation, it leaves your history at once and Anthropic's back-end storage within 30 days. Projects stay until you delete them. If Help improve our AI models is switched on, de-identified chats can stay in training pipelines for up to five years, according to Anthropic's How long do you store my data?

A work or school account follows the retention rules its administrator sets, so ask whoever manages it. Screen recorders raise the same question for everything you see, and our guide to Rewind AI alternatives after the shutdown notes which ones keep that history on your own computer.

ChatGPT: inspect the memory summary and its sources

Open Settings → Personalization → Memory. OpenAI’s current guide describes a memory summary that changes as you chat; it is not a complete inventory of everything influencing personalization. Memory in ChatGPT.

Review the summary and correct an entry through its editing controls. To clear the displayed memories and disable the feature, use the three-dot menu’s Delete and turn off memory option where available. This leaves past chats intact. Re-enabling memory can allow it to build context from conversations that remain.

For fuller removal, delete every relevant source: chats, archived chats, files and information in the memory summary. Disconnect connected apps containing the detail where appropriate. Review custom instructions too.

Older layouts may show saved memories instead. Those are stored separately from chat history, so deleting the original conversation alone does not remove the saved entry. Use the controls your account actually displays; the current help article documents both experiences. If your worry is space rather than privacy, see how much ChatGPT memory can hold and what to do when it fills up.

To change training permission, go to Settings → Data controls and switch off Improve the model for everyone. OpenAI’s Data controls in ChatGPT page explains that conversations can remain in history with training disabled. That setting does not perform the source cleanup above.

For a conversation you do not want in ordinary history, consider Temporary Chat. Check its personalization choice before starting: OpenAI’s Temporary chat page now distinguishes personalized and unpersonalized temporary conversations, and you choose one before you send your first message. The personalized version can reference existing memory without creating new memories. While a chat remains temporary, it is not used to improve models, but OpenAI may retain a safety copy for up to 30 days. Saving it converts it into a regular conversation governed by your account settings.

Are ChatGPT chats private?

Other people cannot see your ChatGPT chats unless you share one, but the chats are not private from OpenAI. On a personal account, OpenAI can use your chats, and what memory keeps from them, to train its models while Improve the model for everyone is switched on.

A shared link is public to anyone who has it. OpenAI says shared pages are not meant for search engines, but that does not make a link private. Links do not expire, and deleting a link removes neither the chat nor copies other people saved. Review yours in Settings → Data controls → Shared links → Manage, as described in Sharing conversations in ChatGPT.

Deleted chats and Temporary Chats are once again removed within 30 days. A court order in the New York Times case made OpenAI keep them indefinitely, and OpenAI says that obligation ended on 26 September 2025. It still stores a limited set of user data from April to September 2025 under that case, according to its update on the data demands. Business, Enterprise and Edu workspaces are not used for training by default.

Gemini: review Memory, instructions and Activity

On the website, open Settings & help → Personal context → Memory. Some Google pages call this section Personal Intelligence. Google’s “Get personalization with memory of your past Gemini chats” documents this switch. The feature requires Keep Activity to be enabled, an eligible personal account and an age of 18 or over.

Turning Memory off stops that personalization feature; it does not itself delete the chats in Gemini Apps Activity. To remove a fact remembered from conversations, Google directs users to delete all chats containing it. Changes may take a little time to affect responses.

Check Your instructions for Gemini on the same Personal context page. Some regions still call it Saved info. Open each entry’s More menu to edit or delete it, or disable instructions at the top. Google’s “Manage what Gemini remembers about you and customize responses” documents these controls.

A request in chat to avoid a subject is not the same as deleting its source. If the information also comes from a connected app, Google says to delete the relevant chats and disconnect that app. Either action alone can leave another route to the information.

In Gemini Apps Activity, review Keep Activity and the deletion controls. Google says that with Keep Activity on, saved chats and shared content can help improve services, including model training, with human reviewers involved. Gemini Apps Privacy Hub.

With Keep Activity off, future chats are not used to train Google’s AI models unless you submit feedback. Google still retains them for 72 hours for service operation, feedback processing and safety. Turning the setting off does not delete existing Activity unless you choose Turn off and delete activity; otherwise review and delete it separately.

Google also says previously reviewed conversations disconnected from your account can remain for up to three years. Activity deletion does not delete information in other Google services. Those limits matter more than an empty recent-chat sidebar.

What Gemini Personal Intelligence can see

Personal Intelligence lets Gemini use data from the Google apps you connect: Gmail, Calendar and Drive, Contacts, Search, Maps and Shopping, your YouTube history, Google Photos in some countries and Google Wallet in the US. Nothing is connected until you opt in, and you choose which apps to connect.

It needs a personal Google Account, an age of 18 or over and Keep Activity switched on. It is not available in the European Economic Area, Nigeria, Switzerland or the UK. Google's Connect your Google apps page lists the current apps and countries.

Google says Gemini does not train its models directly on your Gmail inbox, Drive, Contacts or Calendar. The chats where Gemini used that data are a different matter: with Keep Activity on, they can be used for training, and a sample goes to human reviewers. Google also says Gemini aims to avoid proactive assumptions about sensitive data such as your health, though it will discuss it if you ask.

To switch it off, open Settings & help → Personal Intelligence → Connected Apps and turn off each app. In the Android app, tap Menu, then your profile picture, then Personal Intelligence → Connected Apps. For a single chat, open Tools in the text box and turn Personal Intelligence off; each new chat starts with it on again. Disconnecting an app does not delete anything already saved in Gemini Apps Activity, so delete those chats too if you want the data gone.

Claude: distinguish pausing memory from resetting it

Open Settings → Memory and inspect Topics. You can open an item to read it, edit it or delete it. Anthropic’s “Use Claude’s chat search and memory to build on previous context” documents the current layout; if memory still appears under Settings → Capabilities for you, that is the older layout.

Under Generate memory from chats, disabling memory offers two different choices. Pause memory keeps existing memories but stops their use and the creation of new ones. Reset memory permanently removes memories, including project memories. Resetting memory is not a command to delete your conversation history.

On paid plans, review Search and reference chats separately in the memory settings. Searchable conversations and a stored memory are different ways prior information can become relevant again. Delete source conversations you no longer want retained, and delete the matching memory topic as well: in the current layout, deleting a chat does not remove memory generated from it.

Claude also offers Include sensitive topics in memory. The current guide says these topics are excluded by default unless you enable that setting. That boundary concerns what enters memory; sensitive text you write can still exist in the conversation itself.

For training, open Settings → Privacy and review Help Improve our AI models. Anthropic’s “How do I change my model improvement privacy settings?” explains that switching it off stops previous and new conversations from being used in future model-training runs. It does not undo training already underway or completed, and safety-related exceptions remain.

Anthropic’s “How long do you store my data?” says deleting a conversation removes it from history immediately and ordinarily from backend storage within 30 days, with stated exceptions including legal and safety needs. History disappearance is therefore not confirmation of instant removal everywhere.

For a fresh conversation outside a project, the ghost icon starts an incognito chat. Anthropic’s “Use incognito chats” says these chats do not enter history or memory and are not used for training. They are retained for 30 days by default. Profile preferences can still apply, so incognito does not mean the service knows no context about you.

A privacy cleanup you can finish today

Keep reference storage deliberate too

dEssence offers another place for things you choose to save through its Chrome extension, Telegram bot or web app. That is a workflow choice, not evidence of stronger privacy than an assistant. Before saving sensitive material, inspect the service’s current privacy terms and deletion controls. Keep secrets out of a convenience archive when the task only needs a harmless reference. For private notes you write yourself, an end-to-end encrypted notes app is the stronger choice; compare Standard Notes alternatives with encryption.

Frequently asked questions

Is AI memory a privacy risk?

Yes, once you share sensitive details. Memory reuses details from past chats, so an address or a health detail you mentioned once can come back in a later answer. Use a temporary or incognito chat for anything sensitive, and delete both the chat and the memory entry when you want a detail gone.

How do I turn off AI memory?

In ChatGPT, open Settings → Personalization → Memory and switch it off, or choose Delete and turn off memory in the ••• menu of the memory summary. In Gemini, open Settings & help → Personal context → Memory. In Claude, open Settings → Memory → Generate memory from chats and choose Pause memory or Reset memory. None of these deletes your chats.

Does ChatGPT memory use my data for training?

Yes, while Improve the model for everyone is on. On a personal account, OpenAI may use your chats and remembered information to improve its models. Switch it off in Settings → Data controls. Even then, a thumbs up or thumbs down on an answer means that whole conversation may be used for training.

Does AI memory save sensitive topics?

Claude is the only one with a switch for it. Include sensitive topics in memory, in Settings → Memory, is off by default, so health, religion, politics and similar topics stay out of memory until you turn it on. Google says Gemini may use sensitive details from your chats to personalize answers while Memory is on. OpenAI says sensitive information can appear in ChatGPT memory if you share it.

Can I use an AI assistant without it remembering anything?

Yes. A Temporary Chat in ChatGPT never creates memories and stays out of your history, though OpenAI may keep a copy for up to 30 days. A temporary chat in Gemini is kept for 72 hours and is not used to personalize or train. An incognito chat in Claude stays out of history and memory and is kept for 30 days by default.

Are ChatGPT Chats Private? What AI Memory Stores in ChatGPT, Gemini and Claude