Is it safe to give ChatGPT personal information?
Some of it, but not everything. ChatGPT is a mainstream product from a large company, but OpenAI's own privacy policy says that no internet transmission is fully secure and that "you should take special care in deciding what information you provide to the Services."
That is the right way to think about it. Anything you type or upload can be stored, remembered, reviewed for misuse and, on personal plans by default, used to train OpenAI's models. Here is where your information goes, what to keep out, and the settings that reduce the risk, all from OpenAI's own privacy policy and help pages. For the wider question of how private ChatGPT is overall, see is ChatGPT private.
What ChatGPT collects when you share something
OpenAI's privacy policy says it collects the personal data you put into the service, "including your prompts and other content you upload, such as files, images, audio and video, and data from connected services, depending on the features you use."
On top of what you type, it lists:
- Account details. Name, contact information, date of birth, payment information and transaction history.
- Log data. Your IP address, browser type, the time of each request and how you use the service.
- Location. A general area based on your IP address, or more precise GPS location if you choose to share it.
- Contacts. If you connect your device contacts, OpenAI uploads your address book and checks which of your contacts also use its services.
The policy also says OpenAI may monitor content submitted to its platforms to prevent fraud and misuse, and may share personal data with government authorities when it is legally required to. OpenAI says it does not sell personal data.
Why personal information is risky to share
Training is on by default
OpenAI's data controls page says: "On personal workspaces with ChatGPT Free, Plus, or Pro, data sharing is enabled by default." The privacy policy says OpenAI may use your content "to train the models that power ChatGPT."
OpenAI says it takes steps to reduce personal information in training datasets. That is a reduction, not a guarantee, which is why it is better to keep personal details out in the first place.
Memory keeps details between chats
ChatGPT's Memory feature personalizes answers using what it knows about you. According to OpenAI's Memory help page, that can include past chats, saved memories, custom instructions, saved files and "content from connected apps, such as Gmail."
The same page says plainly: "Sensitive information can appear in Memory if you share it with ChatGPT." Two details matter here:
- Deleting a chat "does not necessarily delete a separate saved memory created from that chat."
- Disconnecting an app "does not delete past conversations that already used app content."
So if you connect Gmail and ask about an email with your bank details in it, removing the connection later does not remove what was already said in those conversations.
Copies are kept for up to 30 days
Deleting is not instant. OpenAI's privacy policy says deleted personal data is removed within 30 days, unless it needs to be kept longer for legal reasons, or it was already de-identified and used to improve models. Temporary Chats may be kept for up to 30 days for safety purposes, and logs of deleted memories may be kept for up to 30 days for safety and debugging.
If OpenAI receives something like a lawful subpoena, the policy says it may keep data for as long as that legal obligation lasts.
Feedback buttons share the whole conversation
This one surprises people. OpenAI's data controls FAQ says: "If you choose to provide feedback, such as selecting thumbs up or thumbs down on a response, the entire conversation associated with that feedback may be used to train OpenAI models." Not just the reply you rated, the whole chat.
GPTs with actions can send data elsewhere
Some custom GPTs use actions to talk to outside services. OpenAI's Temporary Chat help page says data sent to a third party through an action "is governed by the recipient's privacy policy. The recipient may retain that data for longer than 30 days and may use it for other purposes." Once your information leaves through an action, OpenAI's settings no longer apply to it.
The privacy policy says the same about information you share with third-party search and shopping partners: their own terms and privacy policies apply.
Shared chats and work accounts
When you share a ChatGPT conversation, you are handing over everything in it, including anything personal you typed earlier in the chat. Read it before you share it.
If you use ChatGPT through an employer, the privacy policy says administrators of an Enterprise or business account "may access and control your OpenAI account, including being able to access your Content."
What not to share with ChatGPT
Given all of the above, keep these out of any chat you would not want stored, remembered or reviewed:
- Passwords, recovery codes and one-time codes. There is no reason ChatGPT needs them.
- Full ID numbers. Social Security numbers, passport and driver's license numbers.
- Card and bank account numbers. If you want help with a bank statement, remove the account numbers and your address first.
- Health records and test results with your name on them. Describe the situation instead of uploading the document.
- Other people's personal information. Their phone numbers, addresses, messages or photos are not yours to hand over.
- Confidential work material. Your employer's rules come first.
- Your exact home address and date of birth together. Each one alone is common, but combined with your name they identify you.
A useful habit: replace real names and numbers with placeholders. ChatGPT can help you draft a letter to "my landlord at [address]" just as well as one with the real address in it.
A short checklist before you share anything
- Use Temporary Chat, unpersonalized. Open a new chat, select Temporary, then choose Unpersonalized before you send your first message. OpenAI says Unpersonalized "does not use memory, custom instructions, or plugins," temporary chats do not create or update memories, and they are not used to improve its models. You cannot change the setting once the conversation starts.
- Turn off Improve the model for everyone. On the web, open your account menu, select Settings, select Data controls, turn off Improve the model for everyone and select Done. On iOS and Android, open the sidebar, select your profile icon, select Data controls and turn it off. OpenAI says this applies across web and mobile when you are signed in, and covers new conversations. Voice mode has its own separate control.
- Review Memory. Go to Settings, then Personalization, then Memory, and delete anything personal you would rather ChatGPT did not keep.
- Skip the thumbs up and thumbs down on sensitive chats. Feedback can send the entire conversation for training.
- Delete chats you no longer need. Remember that turning off training does not delete chats, and archiving a chat does not change whether it can be used for training. Deletion still takes up to 30 days.
You can also opt out through OpenAI's Privacy Portal by selecting Do not train on my content. For the same setting in other apps, see how to opt out of AI training.
Frequently asked questions
Is it safe to give ChatGPT my email address?
Your account already has your contact information. Typing someone else's email address into a chat is different: it becomes part of that conversation's content, with everything above applying to it.
Is it safe to give ChatGPT access to Gmail?
That is your call, but know what it means. OpenAI's Memory page lists content from connected apps such as Gmail as something ChatGPT can draw on, and disconnecting later does not delete past conversations that used it. If your inbox holds bank, health or legal mail, think twice. See also how to stop AI from reading your emails.
Does turning off training delete my chats?
No. OpenAI's FAQ says turning off Improve the model for everyone does not delete or hide saved chats. You have to delete them separately.
Are ChatGPT Business and Enterprise different?
Yes, for training. OpenAI says that by default it does not use content from ChatGPT Business, Enterprise, Edu or ChatGPT for Healthcare workspaces, or from its API, to train its models. Your company's admins can still access your content.
Can I trust what ChatGPT tells me about my personal situation?
OpenAI's own privacy policy says "you should not rely on the factual accuracy of output from our models." Check anything important with a professional.
Keep the personal stuff in your own inbox
A lot of what people paste into ChatGPT starts in their email: statements, medical letters, contracts. Where that mail lives matters too. Secria Mail has no ads, Secria never mines your mail, and your mailbox is locked with a key made from your password, on your device. Sign-up is a name, the address you want and a password, with no phone number. Create a free account.
Secria fact-checks every post against primary sources. Spotted something wrong or out of date? Email hq@secria.me and we will correct it.