Is ChatGPT safe for confidential information? A plain-English answer
The short answer: on a personal ChatGPT account (Free, Plus or Pro), treat anything you type as not confidential. Your chats are stored until you delete them, they can be used to train OpenAI's models unless you switch that off, and turning training off doesn't change where the data is kept. ChatGPT Business and Enterprise workspaces come with stronger commitments: OpenAI says it doesn't train on that data by default, and admins control retention. Even then, your company's rules, contracts and clients decide what is allowed, not the tool.
So the practical rule is: keep truly confidential details out, and give ChatGPT a placeholder version instead. You'll get almost the same help with far less risk.
This page explains what OpenAI's own documentation says happens to your information, how personal and business plans differ, and a four-step check before you paste anything sensitive. If you want the list of things never to type in, see what not to tell ChatGPT. This guide is general information, not legal advice.
What does "safe" mean for confidential information?
When people ask whether ChatGPT is safe for confidential information, they usually mean four different things. It helps to separate them:
- Will it be used to train the model? Could my text shape future answers?
- How long is it kept? Does it sit on someone else's servers after I've finished?
- Who can see it? Could a person at OpenAI, a colleague or a stranger read it?
- Am I allowed to share it at all? Does a contract, a client or my employer forbid it?
The first three depend on your plan and settings. The fourth doesn't depend on ChatGPT at all, and it's the one people forget.
What happens to what you type on a personal account?
These answers come from OpenAI's Help Center pages, read on 27 October 2026. OpenAI updates them often, so check the linked pages before relying on a detail.
Training: on by default, and you can turn it off
OpenAI's Data controls help page says that on personal workspaces with ChatGPT Free, Plus or Pro, "data sharing is enabled by default". To stop your new conversations being used for training, go to Settings > Data controls and turn off Improve the model for everyone.
Two catches matter for confidential work:
- It only covers new conversations. OpenAI's page on how your data is used to improve model performance says that after you opt out, "we won't use your new conversations to improve our models".
- Feedback overrides it. The same page says that if you give a thumbs up or thumbs down, "the entire conversation associated with that feedback may be used to train our models", even if you've opted out. If a chat contains anything sensitive, don't rate it.
Retention: kept until you delete it
Switching training off doesn't delete anything. OpenAI's chat and file retention page says regular and archived chats "remain saved to your account until you delete them". When you delete a chat, it disappears from your view straight away and OpenAI schedules it for permanent deletion "within 30 days", unless it must be kept longer for security or legal reasons. Deleted chats can't be restored.
Files behave slightly differently. The same page says files in a custom GPT are kept until the GPT is deleted, files in a project until the project or file is deleted, and where Library is available, deleting a chat doesn't delete a file saved there.
Temporary chat is the closest thing to "don't keep this". OpenAI says temporary chats don't appear in your history, don't create memories and aren't used to improve its models, but "may be retained for up to 30 days for safety purposes". So it reduces what's kept; it doesn't make the conversation vanish instantly.
Shared links: anyone with the link can read it
This is the easiest way confidential information leaks, and it has nothing to do with training. OpenAI's shared links help page says that on a personal account, "anyone who has a link" can view the shared conversation. Links don't have expiry dates or per-person access, and OpenAI notes that although shared pages aren't meant for search engines, "this does not make a link private". Anyone who has it can forward it.
You can review and delete your links under Settings > Data controls > Shared links > Manage. Deleting a link stops future access, but it doesn't remove a copy someone has already saved to their own account.
How are ChatGPT Business and Enterprise different?
OpenAI makes a separate set of promises for what it calls business data. Its Enterprise privacy page (updated 8 January 2026) and business data page say:
- No training by default. OpenAI does not use data from ChatGPT Business, Enterprise, Edu, Healthcare, Teachers or the API Platform for training "by default", unless the organisation explicitly opts in. The model-improvement page adds that on Business, Enterprise and Edu, a thumbs up or down "does not permit training on the associated conversation".
- Admins control retention. Workspace admins set how long data is kept. For ChatGPT Business, deleted or unsaved conversations are removed from OpenAI's systems within 30 days, unless the law or harm prevention requires longer.
- Encryption. Data is encrypted at rest (AES-256) and in transit (TLS 1.2 or higher).
- Limited human access. For ChatGPT Business, OpenAI says its access is limited to authorised employees for engineering support, abuse investigations and legal compliance, and to "specialized third-party contractors who are bound by confidentiality and security obligations, solely to review for abuse and misuse". Inside your own company, Business workspace admins "can view, access, export, and delete" members' conversations.
- Shared links stay inside the workspace. Links created in a Business, Enterprise or Edu workspace can only be opened by eligible members of the same workspace.
- Contracts for compliance. OpenAI says it can sign a Data Processing Addendum for ChatGPT Business, Enterprise and the API, and a Business Associate Agreement for ChatGPT for Healthcare and API healthcare customers.
That's a real difference. But read the list again: your admins can see your chats, business data may still be run through automated safety tools, and a business plan can't give you permission to share something a client contract says you mustn't share.
So, should you paste confidential information into ChatGPT?
Use the plan you're on as the first filter:
- Free, Plus or Pro (personal): don't paste confidential information. Use placeholders.
- A Business or Enterprise workspace your employer set up: follow your employer's AI policy. If there isn't one, ask before pasting client, staff or financial data.
- Anything covered by a contract, NDA or professional rules: check what that agreement says about outside services. If you're unsure, the answer is no.
A 4-step check before you paste anything sensitive
- Ask "whose information is this?" If it belongs to a client, a patient, a colleague or a partner, you need their rules, not just yours.
- Check your account. Look at the top of ChatGPT: is this your personal account or your work workspace? Many people have both and paste into the wrong one.
- Strip it down. Replace names, numbers, prices, addresses and anything identifying with placeholders like [Client A], [amount] and [date]. Only paste the part you need help with.
- Tidy up after. Use a temporary chat for one-off sensitive questions, don't rate the answer, don't create a shared link, and delete the chat if you no longer need it.
Copyable prompts that keep the details out
You can get help with confidential work without giving the confidential parts away. Paste these as written, then fill in your own placeholders.
Plan the stripped-down version first (paste nothing sensitive):
I need help with a confidential [type of document, e.g. supplier contract]. Without me sharing it, list the kinds of details I should replace with placeholders before I paste any of it into an AI chat, and suggest a placeholder for each.
Work on the placeholder version:
Here is a clause from a contract. I've replaced the names and figures with placeholders. Explain in plain English what it commits [Party A] to, and list the questions I should ask before signing. [paste clause with placeholders]
Get a reply drafted without the client's details:
A client, [Client A], is unhappy that [project] is [number] days late because [reason in one line]. Draft a short, calm reply that apologises once, gives the new date as [date], and offers [what you can offer]. I'll add the real details myself.
Then put the real names and numbers back in your own document, outside ChatGPT, and check the result before it goes anywhere. Our five-minute check covers what to look for.
Frequently asked questions
Is ChatGPT confidential?
Not in the sense a lawyer or doctor is. On a personal account, your chats are stored on OpenAI's systems until you delete them and can be used for training unless you opt out. Business and Enterprise plans add stronger commitments, but your workspace admins can still access conversations. Treat it as an outside service, not a confidential adviser.
Does turning off "Improve the model for everyone" make ChatGPT private?
No. It stops your new conversations being used for training. OpenAI says it doesn't delete or hide your chats, which stay in your history until you delete them. Rating an answer can still send that whole conversation for training.
Does temporary chat delete everything?
Temporary chats don't appear in your history, don't create memories and aren't used for training. OpenAI says it may keep a copy for up to 30 days for safety purposes, so it's safer, not invisible.
Can my employer see my ChatGPT chats?
In a ChatGPT Business workspace, OpenAI says admins can view, access, export and delete members' conversations. Enterprise admins can access conversation audit logs through OpenAI's Compliance API. On your own personal account, your employer has no access through ChatGPT. That's also why pasting work data into a personal account is usually against company policy.
Is ChatGPT Business safe for client data?
It's built for it: no training by default, admin-controlled retention, encryption and a Data Processing Addendum. Whether you may put a particular client's data in still depends on your contract with that client and the law that applies to you. Check both, and keep identifying details out where you can.
Is it safe to use ChatGPT at work?
Usually yes, with a few rules agreed in advance. Our guide to using ChatGPT safely at work covers the setup, the checking routine and a one-page team policy.
Not ready to buy? Start with the free guide, Your first week with AI: ten prompts and a privacy checklist, plus a five-day email course.