Can GPT creators see your chats? What custom GPTs send out, and what retirement leaves
The Morse teamPublished 8 min read
No, as of November 2026 GPT creators can't see your chats with their custom GPTs in ChatGPT, because OpenAI says builders "cannot view individual conversations that users have with their GPTs" (OpenAI). A GPT with actions is different. When it calls an outside API that the builder set up, "relevant parts of your input may be sent to the third-party service," and OpenAI says it "does not audit or control how those services use or store your data" (OpenAI). So the builder never sees the chat in ChatGPT, but the builder's own server can receive what you typed.
We read OpenAI's help pages on GPTs, sharing and publishing, actions, workspace controls, retention, Temporary Chat, data controls, connected apps, plugins, and the GPT retirement FAQ, along with its developer docs for GPT Actions, its usage policies, and its privacy policy, on November 16, 2026. OpenAI deserves credit here. Its pages say in plain words that builders can't read your chats, and that a service an action calls "may retain that data for longer than 30 days and may use it for other purposes" (OpenAI).
Custom GPT privacy: the key facts
Read November 16, 2026.
| Setting | Default | How long it's kept | Source |
|---|---|---|---|
| Builder reading your chat | Not possible: builders "cannot view" individual conversations | Not applicable | OpenAI |
| Your chat with a GPT | Saved to your ChatGPT history | Until you delete it, then within 30 days, with security and legal exceptions | OpenAI |
| Training on GPT chats | On for personal Free, Plus, and Pro; off by default on Business, Enterprise, and Edu | Follows your plan's training rules | OpenAI, OpenAI |
| Memory, custom instructions, and past chats | Not used by GPTs; "Each conversation starts fresh" | Not applicable | OpenAI |
| What an action sends | Parts of your conversation go to the API the builder defined, sometimes after an approval prompt | Under the recipient's policy, possibly "longer than 30 days" | OpenAI, OpenAI |
| The builder's knowledge files | Used to answer you | Until the GPT is deleted, then within 30 days | OpenAI |
| Custom GPTs themselves | Retiring | December 11, 2026; February 11, 2027 for FedRAMP and approved deferrals | OpenAI |
What can a GPT builder see about you?
Very little inside ChatGPT. A published GPT's public page "may include its name, icon, description, category, capabilities, conversation starters, ratings if available, and builder profile details" (OpenAI). None of that is your conversation. OpenAI's current help pages don't say what usage numbers, if any, a builder sees about the people who use a GPT, so we can't tell you more than that.
The builder profile tells you about the builder instead. It can show a name, "a website link from a verified domain," and "supported verified social profiles," and for personal accounts, verification "can involve billing details or verifying ownership of a public domain name" (OpenAI). Check it before you type anything private into a GPT whose maker you don't know.
What does a GPT action send to the builder's server?
The parts of your conversation the action needs. Actions connect a GPT "to external APIs that you define," in OpenAI's words to builders (OpenAI), and OpenAI's developer docs say "ChatGPT may send parts of their conversation to the action's endpoint" (OpenAI). That endpoint can be the builder's own server.
From there, the builder's policy applies. OpenAI's Temporary Chat page says data sent through an action "is governed by the recipient's privacy policy," and the recipient "may retain that data for longer than 30 days and may use it for other purposes" (OpenAI). Deleting the chat in ChatGPT doesn't reach that copy, as we note in how to delete your ChatGPT history.
Who gets what from one GPT chat
We built this from seven OpenAI pages on November 16, 2026.
| Who | What they get | Whose rules apply | Source |
|---|---|---|---|
| The GPT's builder, in ChatGPT | Not your conversation | OpenAI's | OpenAI |
| The API an action calls | "Relevant parts of your input," up to 100,000 characters of text per request | The recipient's privacy policy | OpenAI, OpenAI |
| That API, if you sign in to it | Your token for that service, on every call | The recipient's privacy policy | OpenAI |
| An app the GPT uses | What your own connected account already allows | The app's permissions | OpenAI, OpenAI |
| OpenAI | The whole chat, saved to your account | OpenAI's privacy policy and your training setting | OpenAI |
| Your workspace admins | Your content, on a business account | Your employer's | OpenAI |
A few details narrow what the API learns. ChatGPT calls an action "from one of the published IP ranges," so the request comes from OpenAI's servers, not your device (OpenAI). Requests "can only contain text (no images or video)" (OpenAI). If you sign in to the service through the GPT with OAuth, "Each time a user makes a request to the action, the user's token will be passed in the Authorization header" (OpenAI), so the service knows which of its accounts is asking.
OpenAI requires a valid privacy policy URL for actions in "Public GPTs (link or GPT Store)" (OpenAI). Its pages state no such rule for a GPT shared only with you or inside a workspace. OpenAI's usage policies, in the version effective November 20, 2026, forbid using its services to "aggregate, monitor, profile, or distribute individuals' private or sensitive information without their authorization" (OpenAI). OpenAI's own advice to users is short: "Only use GPTs with APIs or apps that you trust" (OpenAI).
Does ChatGPT ask before a GPT sends your data?
Sometimes, and the builder decides how often. OpenAI says "ChatGPT may ask you to approve the request before information is sent or the action runs," depending on "the GPT's setup and the type of action" (OpenAI).
The developer docs explain the switch. A builder can mark an endpoint with x-openai-isConsequential: true, and ChatGPT will then "always prompt the user for confirmation before running" and won't show an "always allow" button (OpenAI). Set to false, ChatGPT "shows the 'always allow button'." If the builder sets nothing, "ChatGPT defaults all GET operations to false and all other operations to true." So a lookup, the kind of request that fetches something for you, offers Always allow by default, while one that changes something asks every time. Skip Always allow for any GPT whose builder you haven't checked.
A GPT can use apps or actions, "but not both at the same time" (OpenAI). Apps run on your own connected accounts, and how to cut one off is in how to disconnect apps from ChatGPT.
Do your chats with a custom GPT train OpenAI's models?
They can, depending on your plan rather than the builder's. OpenAI says that on Business, Enterprise, and Edu, "data is not used for training by default," while on consumer plans it "may be used for training depending on whether you have opted out" (OpenAI). On personal Free, Plus, and Pro, "data sharing is enabled by default" (OpenAI).
To turn it off on the web, open your account menu, select Settings, then Data controls, and turn off Improve the model for everyone (OpenAI). The exceptions to that switch, such as rating a reply, are in does ChatGPT train on your data.
Who else can see your GPT chats at work?
Your workspace's administrators. OpenAI's privacy policy says business account administrators "may access and control your OpenAI account, including being able to access your Content" (OpenAI). What else they can reach is in is ChatGPT Business private.
Workspaces do get one control that personal accounts lack. In Enterprise and Edu, admins can allow third-party GPTs, only owner-approved ones, or none, and can limit actions in GPTs made in the workspace to approved domains: "If no domains are added, GPT actions are not allowed" (OpenAI).
What happens to your GPT chats when custom GPTs retire?
They stay in your history. OpenAI says "Your existing conversations with custom GPTs will remain accessible after retirement," even though GPTs "and their GPT pages become inaccessible" on December 11, 2026, or February 11, 2027 for FedRAMP workspaces and approved deferrals (OpenAI). Those chats sit on OpenAI's servers until you delete them, then go "within 30 days," unless they were already de-identified or OpenAI must keep them for security or legal reasons (OpenAI).
Personal accounts already can't make new ones, though "Existing GPTs remain available to use." "New GPT creation and publishing are not available on personal ChatGPT accounts, including Free, Go, Plus, and Pro," and OpenAI points people to Plugin Creator for new workflows (OpenAI). Affected Enterprise workspaces lost GPT creation on October 26, 2026 (OpenAI).
Migration moves what the builder made. A GPT's instructions "become a skill within the new plugin," its knowledge files "are copied into the plugin's reference files," and the plugin "starts private" (OpenAI). Custom actions "do not transfer," so a GPT's API connection has to be rebuilt before the plugin can use it.
Plugins behave differently from GPTs. "ChatGPT automatically uses your installed plugins when they're relevant to your request" (OpenAI), where a GPT works only when you open it or mention it. OpenAI's plugin page adds that its OpenAI Verified badge "does not replace your organization's privacy, security, or vendor review."
How Morse keeps your chat history in your browser
A GPT chat runs on OpenAI's servers and is saved to your account, and a builder's API sometimes keeps its own copy under its own policy. Retiring GPTs leaves those chats in place until you delete them.
On Morse, your chats are stored in your browser, encrypted, not on Morse's servers, and characters you make are stored the same way. Our servers don't keep your messages or log what you write. Morse doesn't train AI on your chats, sell your data, or show ads, and because we don't store chats, we usually have very little to give authorities. Our privacy policy lists what the account holds: sign-in details, plan, balance, and a usage ledger with no message content.
Pick the mode for what you'd have typed into a GPT:
- Private, the default, for the everyday questions you'd have taken to a GPT. Open-weight models like DeepSeek, Kimi, and GLM, plus xAI's Grok, answer with your account, email, and IP address removed. Morse has zero data retention agreements with its inference, search, and other providers, so they've agreed to keep nothing, and where a provider confirms zero retention on each reply, Morse checks it and refuses a reply that doesn't.
- E2EE, for anything you want only a verified enclave to read. Your browser checks the secure enclave's proof that it's genuine before anything leaves your device, then encrypts the message so only that enclave can read it, and our servers relay ciphertext they can't read. The enclaves are run by third-party providers, not by Morse. E2EE runs open-weight models with no web search or image generation, and it rests on the enclave hardware and attestation working as designed.
- Anonymous, to use GPT, Claude, and Gemini without the provider knowing it's you. Morse's proxy removes your account, email, and IP address and sends the message under Morse's own account. The provider sees the message without knowing it's you and may keep it for a limited time to check for abuse, under its own policy, so leave out names and details that point to you.
Clearing your browser, losing the device, or deleting your account deletes your chats, and Morse can't recover them, so save a copy of anything you need.
Try a chat from the Morse home page with no account, or start free; E2EE chats come with Pro, at $19 a month.
Common questions
Can a GPT builder see my name or IP address?
Not through the chat. OpenAI says builders can't view your conversations, and ChatGPT calls a GPT's actions from OpenAI's published IP ranges, not from your device. If you sign in to the builder's service through the GPT, every call after that carries your token for that account, so the service knows which of its users is asking.
Can I still make a custom GPT with ChatGPT Plus?
No. As of November 2026, OpenAI says new GPT creation and publishing aren't available on personal accounts, including Free, Go, Plus, and Pro. GPTs you already made still work, and OpenAI points new projects to Plugin Creator.
Can a custom GPT see my other ChatGPT chats or my memories?
No. OpenAI says GPTs don't use saved memory, custom instructions, or previous conversations, and each GPT conversation starts fresh.
What happens to a GPT's knowledge files when it moves to a plugin?
OpenAI says migration copies them into the plugin's reference files, and the new plugin starts private. Files on a GPT are kept until the GPT is deleted, then scheduled for deletion within 30 days.