Is ChatGPT private? Memory, shared links, staff access, and what deleting leaves

The Morse teamPublished October 25, 202611 min read

As of October 2026, ChatGPT is private from other people unless you share a link, but it isn't private from OpenAI. OpenAI keeps your chats on its servers until you delete them, builds memories from them, lets a limited number of staff and outside service providers read them for safety, support, legal matters, and model improvement, and shares data with government authorities when the law requires it. Temporary Chats are still kept for up to 30 days, and deleting a chat doesn't delete the memories ChatGPT saved from it.

We read OpenAI's US privacy policy and its help pages on data controls, memory, Temporary Chat, chat retention, shared links, connected apps, data export, and account deletion on October 20, 2026. OpenAI deserves credit for how plainly those pages now read. The help center says in writing what deleting a chat leaves behind.

ChatGPT privacy at a glance

Can people see my chats on ChatGPT?

No, not unless you share them. Other ChatGPT users can't browse your history. The exception is a shared link, and OpenAI's shared links page is blunt about it: "Anyone who has a link created from a personal account can view the shared conversation."

Before you press Share, know three things from that page:

• There's no audience control. OpenAI says "Shared links do not offer recipient-by-recipient access controls," and links have no "configurable expiration dates."
• Links travel. Shared pages "are not intended for search-engine indexing, but this does not make a link private. Anyone who can access a link can also forward it."
• Files and images go too. "Supported images and uploaded files can appear in a shared conversation."

Search engines did once index ChatGPT chats. In July 2025, more than 4,500 shared links turned up on Google after users ticked an opt-in box to make them discoverable. OpenAI removed the option on July 31, 2025, saying it "introduced too many opportunities for folks to accidentally share things they didn't intend to" (Search Engine Journal).

Who at OpenAI can read your ChatGPT chats?

Some staff and some outside companies, for four stated reasons. OpenAI's consumer data page says "a limited number of authorized OpenAI personnel, as well as trusted service providers" may access your content for:

1. "investigating abuse or a security incident,"
2. support, if you contact OpenAI or a product issue needs troubleshooting,
3. "to handle legal matters," and
4. "to improve model performance (unless you have opted out)."

OpenAI says it logs "all access to user content," limits it "on a need-to-know basis," and trains staff before they get it. The same page also says OpenAI "may also send select portions of content to trusted service providers" for "data annotation and safety purposes," and ends with a plain request: "Please do not enter sensitive information that you would not want reviewed or used."

Turning off training removes only the fourth reason. Abuse checks, support, and legal matters stay.

The annotation work is where the Project Lily lawsuit sits. Two users allege that contractors read real chats, sometimes with a summary of the user's memories; those claims are unproven, and we go through the filing in ChatGPT Project Lily.

What does ChatGPT remember about you?

More than your chat list shows. OpenAI's memory page says Memory can draw on "past chats," "saved memories," "custom instructions," "files in Library," and "content from connected apps, such as Gmail," depending on your plan and region. The controls are under Settings, Personalization, then Memory, where you may see Reference saved memories, Reference chat history, and a Memory summary.

Each source has to be deleted on its own: deleting a chat doesn't delete the memory saved from it, deleting a memory doesn't remove it from past chats, "Don't mention this again" deletes nothing, and turning off Reference chat history schedules what it learned for deletion within 30 days while your chats stay. The Memory summary won't show you all of it either, since OpenAI says it "does not necessarily include every detail or source that Memory can reference."

What you tell ChatGPT can also travel past the chat window. With Memory on, OpenAI says ChatGPT "may use relevant details to formulate a more useful search query," such as your location or diet, and its privacy policy says information you share with third-party search and shopping partners "is governed by their own terms and privacy policies" (OpenAI).

OpenAI's data controls also say turning off Memory "does not disable safety features that may use limited, safety-relevant context in rare, high-risk situations." In May 2026, OpenAI described these as "safety summaries," notes about earlier chats involving suicide, self-harm, or harm to others that are "kept only for a limited time" (OpenAI). It doesn't say how long.

Adults can also add one trusted contact under Settings, then Trusted contact. If automated systems and trained reviewers find you may be talking about suicide in a way that signals a serious safety concern, ChatGPT may notify that person and tell them only the general reason: "Your chat details or transcripts are not shared" (OpenAI).

Is ChatGPT's Temporary Chat private?

It's more private than a regular chat, but "OpenAI may keep a copy of a temporary chat for up to 30 days for safety purposes" (OpenAI), and a Personalized one can read your memories, so pick Unpersonalized before your first message. What it means for training is in does ChatGPT train on your data, and what it hides about your account is in how to use ChatGPT anonymously.

What do connected apps see in ChatGPT?

More than the file you asked about. OpenAI's connected apps page says an enabled app "may be able to access information from ChatGPT," including "relevant context from your ChatGPT conversations," and, with Memory on, "relevant context from your memories." Apps "may also see basic information," such as your "IP address, device or browser type, language and region settings, and approximate location."

After that, the app's own rules apply: "Data shared with apps is handled according to each app's terms of service and privacy policies." On Free, Go, Plus, and Pro, OpenAI "may use information accessed from apps to train our models" if "Improve the model for everyone" is on. Disconnecting an app stops future access but "does not delete past conversations that already used app content." You manage apps under Settings, then Plugins.

Who does OpenAI share your ChatGPT data with?

OpenAI's privacy policy, section 3, names seven kinds of recipient. We put them in one table with the condition the policy attaches to each.

Government requests for chat content nearly doubled in six months. OpenAI's transparency report for January to June 2026 counts 146 content requests, with data disclosed in 88 of them, covering 130 accounts, plus 250 to 499 national security requests. Its previous report, for July to December 2025, counted 75. Courts can reach your chats too: a preservation order in the New York Times case kept even deleted chats until September 26, 2025, which we cover in the ChatGPT court order on data retention.

Advertisers aren't on the list. OpenAI says it "does not sell your data or share your ChatGPT conversations with advertisers," although on Free and Go your chats help choose the ads you see (how to turn off ChatGPT ads).

Is ChatGPT private if you're not logged in?

It's more private from your name, not from OpenAI. Signed out, "you can only have one conversation at a time," and "Chats will remain accessible only for the current session on the same browser" (OpenAI). The same page says signed-out use isn't available in Europe, except in the UK. Training, ads, the device details OpenAI still collects, and how long it keeps these chats are covered in can you use ChatGPT without an account.

Is ChatGPT confidential, and is it safe for confidential information?

Not in the legal sense, and OpenAI's chief executive has said so. In July 2025, Sam Altman said that with a therapist, lawyer, or doctor "there's legal privilege for it," and "we haven't figured that out yet for when you talk to ChatGPT" (TechCrunch).

Technically, OpenAI says "Your content is encrypted at rest and in transit between you and OpenAI, and between OpenAI and its service providers" (OpenAI). That protects your chat from outsiders on the network. It doesn't stop OpenAI's own systems, which read every message in order to answer it, and OpenAI doesn't describe ChatGPT as end-to-end encrypted.

At work, a managed account is less private from your employer. OpenAI's managed account page says an administrator "may be able to access, export, audit, retain, delete" your prompts, files, and history. Business and Enterprise content isn't used for training by default, which we cover with the rest of training in does ChatGPT train on your data. This isn't legal advice: if a document is privileged, ask your lawyer before you paste it into any AI.

Can ChatGPT leak your chats?

It has happened. On March 20, 2023, a bug in an open-source library "allowed some users to see titles from another active user's chat history," and possibly the first message of a new chat, according to OpenAI's write-up. The same bug may have shown payment details, though not full card numbers, of 1.2% of Plus subscribers active in a nine-hour window.

A later incident didn't touch chats. In November 2025, an attacker exported analytics data from Mixpanel, a vendor OpenAI used on its API site, including names and email addresses of some API users and of some ChatGPT users who had filed help tickets. OpenAI says "No chat, API requests, API usage data, passwords, credentials, API keys, payment details, or government IDs were compromised," and it stopped using Mixpanel (OpenAI).

What does deleting your ChatGPT data leave behind?

Deleting a chat leaves the memories saved from it and any files you saved to Library, and deleting a shared link "does not remove a conversation or task another person already saved" (OpenAI). Deleting your account removes your data "within 30 days," apart from "a limited set of data" kept "where required or permitted by law" (OpenAI).

How Morse keeps your chats off OpenAI's servers

Every row in the tables above exists because ChatGPT keeps your chats, memories, and files on OpenAI's servers, tied to an account with your email and payment details. Morse doesn't keep that copy.

• No server history to share, review, or hand over. Your chat history is stored in your browser, encrypted, and locked when you sign out. Morse's servers don't keep your messages or log what you write, so there's no archive for staff to open, and Morse usually has very little to give if a court asks.
• No training switch, no ad system. Morse doesn't train AI on your content, sell your data, or show ads.
• Deleting means deleted. Your history is stored only in your browser, with no separate memory left on a server. Clearing your browser, losing the device, or deleting your account deletes your chats, and Morse can't recover them, so keep anything you need.
• Temporary means memory only. Temporary chats on Morse are kept in memory only, and attached files are sent with the message and never stored by Morse.

Then pick the mode for what the chat holds:

• Anonymous, to keep using GPT without OpenAI knowing it's you. Morse's proxy removes your account, email, and IP address and sends the message under Morse's own account. OpenAI still sees the words and may keep them for a limited time to check for abuse, under its own policy, so leave out names and details that point to you.
• E2EE, for anything you'd want kept confidential. Your browser checks the secure enclave's proof that it's genuine, then seals the message so only that enclave can read it. Morse's servers relay ciphertext they can't read. It runs open-weight models, and it rests on the enclave hardware and attestation working as designed. Here's how end-to-end encrypted AI chat works.
• Private, the default, for chats no host should keep. Open-weight models like DeepSeek, Kimi, and GLM, plus xAI's Grok, run on hosts that agreed to keep nothing, with your identity stripped as in Anonymous.

Our privacy policy sets out what the Morse account does hold: sign-in details, plan, balance, and a usage ledger with no chat content.

So is ChatGPT private enough for you?

• Other people: yes, if you don't share links. Check Shared links in Data controls now and then.
• OpenAI's reviewers: turn off "Improve the model for everyone" and skip the thumbs buttons, since rating a reply can put the whole chat in training, knowing abuse, support, and legal access remain. In Morse's Anonymous mode, OpenAI gets the message without your account behind it.
• Your own history and memories: delete memories and Library files as well as chats, and use an Unpersonalized temporary chat for one-off questions. On Morse, your history is stored only in your browser.
• Confidential work or legal matters: keep it out of ChatGPT. Start the chat in E2EE on Morse.

You can send a first message from Morse's home page without an account, or start free. End-to-end encrypted chats come with Pro, at $19 a month.