ChatGPT privacy: what OpenAI collects, trains on and keeps, and where the controls are

If client material goes through ChatGPT at work, three questions matter: whether OpenAI trains on it, what it keeps and for how long, and what you can see and switch off. This guide answers each from OpenAI's own pages, quoted and dated, including the plan differences and the two settings screens most people never open. It gives no verdict on whether ChatGPT is private enough for your work; that stays with you and your firm's policy.

Short answer: on a personal plan, OpenAI may train on what you type unless you turn one setting off; on a Business or Enterprise plan it says it doesn't by default; and what ChatGPT keeps and remembers is governed by two more controls that the training switch doesn't touch. Everything below is OpenAI's wording, read on its help center, Privacy Policy and Privacy Portal on 27 October 2026, with the page named each time. Where OpenAI is silent, the page says so.

This is the data question: what goes in, what OpenAI does with it, what you can get back out. The neighbouring question, who else can open a chat, has its own page: is ChatGPT confidential? Equerry, which publishes both, makes a personal assistant for client work and comes up only at the end.

Does ChatGPT train on your data? OpenAI's answer, by plan

OpenAI's article "How your data is used to improve model performance" draws the line at the plan. For individuals: "When you use our services for individuals (Free, Plus, Pro), we may use your content to train our models. You can choose whether your conversations help improve our models." For work accounts: "By default, we don't use inputs or outputs from ChatGPT Business, ChatGPT Enterprise, ChatGPT Edu, or our API to improve our models." Its data-controls article says what the default is on the personal side: "On personal workspaces with ChatGPT Free, Plus, or Pro, data sharing is enabled by default."

Four footnotes to the personal-plan row, all OpenAI's. First, the opt-out covers what comes next: "After you opt out, we won't use your new conversations to improve our models." Second, feedback reopens a chat: "If you choose to provide feedback, the entire conversation associated with that feedback may be used to train our models," opt-out or not; on Business, Enterprise and Edu the same thumbs up or down "does not permit training on the associated conversation." Third, two things have their own switch: "Voice mode has a separate control for sharing data for model training," and "Conversations with OpenAI Support may be used to improve OpenAI services, including our models, if Improve the model for everyone is enabled in Settings." Fourth, the opt-out changes nothing about storage: "Turning off Improve the model for everyone does not delete or hide saved chats."

On what "training" means for the content itself, OpenAI's consumer data article says: "depending on a user's settings, we may use the user's prompts, the model's responses, and other content such as images and files to improve model performance," and the training article adds: "Before using eligible user content to improve our models, we take steps to reduce the amount of personal information in training datasets." The Privacy Portal describes its request as a way "to permanently opt out of your content being used for model improvements."

What ChatGPT collects beyond the chat

The "does it train on my data" question is the one people search, but OpenAI's Privacy Policy (the version for people outside the US, updated February 6, 2026) describes a wider set of data than the conversation. Its own category names, with its examples:

• Account Information: "your name, contact information, account credentials, date of birth, payment information, and transaction history."
• User Content: "your prompts and other content you upload, such as files, images, audio and video, Sora characters, and data from connected services, depending on the features you use."
• Communication Information when you write to OpenAI, and Contact Data if you connect your device's address book: "we upload information from your device address books and check which of your contacts also use our Services."
• Log Data: "your Internet Protocol address, browser type and settings, the date and time of your request, and how you interact with our Services."
• Usage Data: "the types of content that you view or engage with, the features you use and the actions you take, when you submit feedback to a model response, the people with whom you interact," plus time zone, country, device type and connection.
• Device Information, Location Information ("the general area from which your device accesses our Services based on information like its IP address," and more precise GPS location only if you choose to share it) and Cookies and Similar Technologies.

The policy's list of purposes starts with "To provide, analyse, and maintain our Services," and includes "To improve and develop our Services and conduct research," "To personalize and customize your experience across our Services," "To prevent fraud, illegal activity, or misuses of our Services, and to protect the security of our systems and Services, including by monitoring any Content submitted or exchanged on our platforms," and "To comply with legal obligations." It then ties back to training: "we may use Content you provide us to improve our Services, for example to train the models that power ChatGPT." One scope note from the policy itself: it "does not apply to content that we process on behalf of customers of our business offerings, such as our API," which is governed by the customer agreement instead.

On selling and advertising, OpenAI's consumer data article says: "OpenAI does not sell your data or share your ChatGPT conversations with advertisers." Ads themselves are a 2026 development: OpenAI's "Ads in ChatGPT" article says "Ad testing started in the United States on February 9, 2026," that "Ads may appear for users on the Free and Go plans," and that "Plus, Pro, Business, Enterprise, and Edu accounts will not have ads." Where ads personalization is on, "If memory is on, ChatGPT may save and use memories and reference recent chats when selecting an ad," with controls under Settings > Ad Controls. Which plans show them, and what the controls do, is covered in does ChatGPT have ads?

What ChatGPT keeps, for how long, and what memory stores

OpenAI's retention rules depend, in its words, on "the type of chat or file, the action you take, and any workspace retention policy that applies." From the retention, memory and data-controls articles, read 27 October 2026:

• Chats you keep stay. "Regular and archived chats remain saved to your account until you delete them or an applicable workspace retention policy removes them." Archiving "does not change its retention period," and "Archiving a chat does not change whether it can be used to train OpenAI models."
• Deleting is a 30-day schedule, with exceptions. OpenAI "schedules it for permanent deletion from its systems within 30 days, unless the chat was already de-identified and disassociated from your account or OpenAI must retain it longer for security or legal obligations." The Privacy Policy's Retention section says the same of anything you delete, and names the carve-outs: content or accounts banned for policy violations, legal holds such as "a lawful subpoena," and financial record-keeping.
• Temporary chats are the one chat type deleted on a timer. The Privacy Policy: "Temporary Chats will be automatically deleted within 30 days (unless we have to retain them for safety or legal reasons, as described further below)." While temporary they "do not appear in your chat history," "do not create or update memories" and "are not used to improve OpenAI models." Save one and "it becomes a regular chat in your history and follows your account's personalization and model-improvement settings." What a personalized temporary chat can still read is in our Temporary Chat guide.
• Files have their own life. "Deleting a chat does not delete a file that remains saved in Library."
• Memory is a third store. OpenAI's Memory article lists what it can draw on, "Depending on your plan and region": past chats; saved memories; custom instructions; files in Library; and content from connected apps, "such as Gmail." It also says: "Deleting a chat alone does not necessarily delete a separate saved memory created from that chat," and "OpenAI may retain logs of deleted saved memories for up to 30 days for safety and debugging purposes." Turning Reference chat history off means "information remembered from past chats is scheduled for deletion from OpenAI systems within 30 days. The original chats remain in your history unless you delete them."

What you can see of memory is, in OpenAI's words, "a high-level view of information that ChatGPT may use to personalize responses. It does not necessarily include every detail or source that Memory can reference." Its FAQ puts the same point as a plain no: asked whether the summary includes everything ChatGPT remembers, "No. The summary is a high-level view and may omit details that are less relevant, unsuitable for the summary, or derived from sources that are not displayed individually." Two other ways to look, both OpenAI's: "You can also ask ChatGPT what it remembers about you," and the Sources that "may appear below the response" when personal context was used, which "may not show every factor that shaped a response." Correcting and deleting from that screen, on either memory system, is walked through in how to delete ChatGPT memory; what deletion leaves behind is in does ChatGPT remember deleted chats?

Where the controls are: the settings screen, the export, the portal

OpenAI puts the privacy controls in three places, and the one most people never open is the one that shows the most. Labels below are as OpenAI's help articles print them on 27 October 2026; OpenAI renames things, so your own screen wins.

1. Settings in ChatGPT. OpenAI: "Open Settings and select Data controls to manage the options available to your account." That screen holds Improve the model for everyone, Shared links, Export data and account deletion. Memory lives one section over: "Memory settings are in Settings > Personalization." OpenAI's data-controls article is explicit that these are different levers: "Memory controls how ChatGPT personalizes responses. Model-training controls determine whether eligible content can be used to improve OpenAI models. These are separate settings." A newer front door, the Privacy Center inside ChatGPT, "is rolling out to signed-in ChatGPT Free, Go, Plus, Pro, and Business users" and "does not include ChatGPT Enterprise, Edu, or ChatGPT for Healthcare"; on the web it is under your account menu, Help, then Privacy center, and it groups the topics as Personalized chats & ads, Chat privacy and Data use & access. OpenAI says it "explains what each one does and helps you find the setting you need," and that opening it "does not change your settings, delete your data, or override your organization's policies."

2. The export: the closest thing to an audit. If the question is "what does OpenAI actually hold on me," the memory summary is the wrong place to look, by OpenAI's own description above. The export is wider. OpenAI's steps on the web: sign in, "Open your profile menu," "Select Settings," "Select Data controls," then "Under Export data, select Export" and "On the confirmation screen, select Confirm export." On iOS: open your account settings, select Data controls, select Export data, then Confirm export. OpenAI says "ChatGPT sends an email or SMS message when the export is ready. Exports can take up to 7 days to arrive," "The download link expires 24 hours after you receive it," and "The downloaded ZIP file includes your chat history and other relevant account data." Its own warning: "Downloaded data can include sensitive account information. Keep your download link and exported files secure." The export is "Not available for ChatGPT Business, Enterprise, or ChatGPT for Healthcare workspaces"; there, "contact your workspace owner about organization-managed data access."

3. The Privacy Portal, outside ChatGPT. OpenAI runs a separate site, privacy.openai.com, for formal requests, and says it "is intended for" users of its consumer services and "People whose personal data appears in ChatGPT responses." It "does NOT apply to OpenAI's business offerings including ChatGPT Enterprise, ChatGPT Business, ChatGPT Edu and our API platform"; for an account provided by your employer, "your privacy requests for this data should be directed to them, not to OpenAI." The portal page (Last Updated: April 29, 2026) lists five request types: Download my personal data ("a copy of your personal data such as your account details, files you uploaded, files you generated, ChatGPT conversations, logs, and payment information, if added"), Do not train on my content, Delete my ChatGPT account, Delete my GPTs, and Remove my personal data from ChatGPT responses, which you can file "whether or not you have a ChatGPT account." You start with Make a Privacy Request, verify your account, then choose the type; OpenAI's export article calls the first option Download my data at that step. The Privacy Policy adds a mailbox for the same rights, dsar@openai.com, and a data protection officer at dpo@openai.com.

ChatGPT Business and Enterprise privacy: what a work plan changes

If you are searching for "ChatGPT Team privacy," start with a name change: OpenAI's Business FAQ says "ChatGPT Team was renamed to ChatGPT Business on August 29, 2025," and "The rename itself did not change the plan's features, pricing, or usage limits." OpenAI's "Enterprise privacy at OpenAI" page (marked updated January 8, 2026) covers "inputs and outputs from ChatGPT Business, ChatGPT Enterprise, ChatGPT for Healthcare, ChatGPT Edu, ChatGPT for Teachers and our API Platform." Its headline commitments, verbatim: "We do not train our models on your data by default"; "You own your inputs and outputs (where allowed by law)"; and "You control how long your data is retained (ChatGPT Enterprise, ChatGPT for Healthcare, ChatGPT Edu)." It describes its security as "Data encryption at rest (AES-256) and in transit between our customers and us, and between us and our service providers (TLS 1.2+)," and says each of ChatGPT Business, Enterprise, Edu and Healthcare "successfully completed a SOC 2 Type 2 audit"; those are OpenAI's statements about its own controls, which this page reports and does not assess.

Four differences a professional will feel, each from OpenAI's pages read 27 October 2026:

• Retention moves to your firm. Business: "Your workspace admins can control how long your data is retained. Any deleted or unsaved conversations are removed from our systems within 30 days, unless longer retention is required by law, or is reasonably necessary to protect our services or any third party from harm." Enterprise, Edu and Healthcare: "Your workspace admins control how long your data is retained. Any deleted conversations are removed from our systems within 30 days, unless we are legally required to retain them."
• The consumer tools close. Self-service export is "Not available for ChatGPT Business, Enterprise, or ChatGPT for Healthcare workspaces"; account deletion "is not available in managed workspaces and should be directed to the workspace owner"; and the Privacy Portal "does NOT apply" to business offerings. The in-app Privacy Center reaches Business but "does not include ChatGPT Enterprise, Edu, or ChatGPT for Healthcare."
• OpenAI's own access narrows, in its description. For Business: access "is limited to (1) authorized employees that require access for engineering support, investigating potential platform abuse, and legal compliance and (2) specialized third-party contractors who are bound by confidentiality and security obligations, solely to review for abuse and misuse." OpenAI also says business data may run "through automated content classifiers and safety tools," producing metadata that does "not contain any of the business data itself."
• Your organization's access widens. That is the other half of the trade, and it belongs to the "who can see it" question, which is ChatGPT confidential? sets out plan by plan. One line from OpenAI's data-controls article applies to every work plan: "Your organization's privacy, retention, and access policies still apply."

OpenAI says it can "execute a Data Processing Addendum (DPA) with customers for their use of ChatGPT Business, ChatGPT Enterprise, and the API," and that for the API it may "securely retain API inputs and outputs for up to 30 days to provide the services and to identify abuse," with zero data retention available "for eligible endpoints if you have a qualifying use-case." Whether any of this meets a given client's terms, a professional rule or a data-protection law is a judgment for your firm and its advisers, not for this page.

What people at work choose not to paste

Nothing in OpenAI's terms tells you what belongs in a chat; your firm's policy and your client's terms do, and this page doesn't add a rule. One habit among professionals who handle other people's information is to run three checks before the paste, and they map onto the controls above:

• Which account am I in? A personal Plus account and a firm's Business workspace carry different training defaults, different retention and different owners of the privacy request, as every section above shows. OpenAI's line: "Available controls depend on whether you are signed in, your plan, and your workspace settings."
• Is this chat going to train, be remembered, or be kept? Three settings, not one. One habit is to turn Improve the model for everyone off once, use a non-personalized temporary chat for anything one-off, and treat memory as a store to review rather than a log, because OpenAI itself says the summary may not include everything.
• Whose information is this? Material received under an engagement letter, an NDA or a data-room agreement is governed by those terms, whatever the setting says. A common habit is to strip the identifying parts and keep the question: the structure of a fee dispute travels without the client's name, account numbers or the attachment they sent you. OpenAI's own request to consumer users is similar in spirit: "Please do not enter sensitive information that you would not want reviewed or used."

And one check most people skip: what is already in there. The export described above answers it; the memory summary, by OpenAI's description, doesn't fully.

What this page doesn't do

It doesn't say whether ChatGPT is private, safe, secure or confidential, for anyone. It doesn't assess OpenAI's security, certify anything, or interpret privilege, a conduct rule, HIPAA, GDPR or a client contract. It reports what OpenAI publishes as of 27 October 2026, and OpenAI changes its products and pages often: where its current page disagrees with this one, OpenAI's page is right. It also reads the Privacy Policy version for people outside the US (updated February 6, 2026); OpenAI publishes a separate version for US residents, marked updated August 25, 2026, whose category names differ from the ones listed above, and an addendum for Korea.

What Equerry reads, and what you can see and delete

Equerry is a personal assistant who helps you get things done, and it is built around the narrow end of the question this page has been asking. Nothing reaches it that you didn't send: an email thread you CC or forward to your own Equerry address; a PDF, an image, a link or text shared from your iPhone's share sheet; a note you type; a voice note you record in Equerry. If you allow it, it reads your calendar, read-only. It never asks for your inbox, and the original stays where it was. So the privacy question your firm has to answer is about the items you chose to send, one at a time, rather than about an account.

Here is the same audit question asked of Equerry:

Replies that keep the Equerry address on the thread keep arriving, so the client's Thursday reply is there before Friday's call; anyone on the thread can see the address on the CC line. To make what you send searchable and answerable, Equerry processes it, and some of that processing is done by the outside providers our AI disclosure page names; that page says what each one receives, and states that we never train AI models on what you save. There are no ads in Equerry. It takes no action in the world for you, so no message goes out, no booking is made and no purchase is placed, and it makes no claim about privilege, any professional rule, or whether what you send it is confidential: read our privacy policy the way this guide reads OpenAI's, against your firm's policy and your client's terms.

Weighing the same questions for another assistant? Claude data privacy and Copilot privacy read Anthropic's and Microsoft's terms the same way.

Sources, all read 2026-10-27 as rendered pages. OpenAI Help Center: How your data is used to improve model performance (article 5722486); Data controls in ChatGPT (7730893); How OpenAI handles data in consumer services (7039943); Memory in ChatGPT (8590148); Chat and file retention in ChatGPT (8983778); Exporting your ChatGPT history and data (7260999); Privacy Center in ChatGPT (20001488); Ads in ChatGPT (20001047); ChatGPT Business: General FAQ (8542115). OpenAI: Privacy Policy (version for individuals outside the US, marked Updated: February 6, 2026); Enterprise privacy at OpenAI (marked Updated: January 8, 2026); OpenAI Privacy Portal (marked Last Updated: April 29, 2026). All quoted wording is OpenAI's; where its pages are silent, this guide says so rather than filling the gap. Where OpenAI's current page disagrees with this one, OpenAI's page is right. This guide is general information, not legal or compliance advice.

Last updated 2026-10-28.

All product names, logos, and brands are property of their respective owners. Equerry is not affiliated with, endorsed by, or sponsored by any other product mentioned; comparisons are provided for informational purposes.