How to stop Grok from training on your data: every switch on X, Grok.com, and the app
The Morse teamPublished 7 min read
To stop Grok from training on your data as of November 2026, you need two switches, because X and Grok.com each have their own say over how your words are used. On X, go to Settings and privacy, Privacy & safety, Data sharing and personalization, Grok & Third-party Collaborators, and clear the box under Data Sharing. On Grok.com, go to Settings, Data, and turn off Improve the Model (in the Grok app, Settings, Data Controls, Improve the model). Even with both off, a thumbs up or down can put a chat back into training, protected posts can still be copied, and nothing on either site stops X's own license to your posts or takes back what a model already learned.
We read X's help page on Grok (from an archived copy dated November 10, 2026, because help.x.com blocks our requests), X's privacy policy and terms of service, its page on protected posts, the consumer FAQ, privacy policy (last updated October 27, 2026), and Europe privacy addendum of SpaceXSI (formerly xAI), and the Irish Data Protection Commission's statements on Grok, all on November 14, 2026. X's settings still say xAI. Both companies deserve credit for naming each switch on their own pages and spelling out its gaps. Several guides, such as Windows Central's, Tom's Guide's, BGR's, Redact's, and Roboin's, give only X's box.
What does each Grok training switch leave out?
Every switch we found on X's and SpaceXSI's own pages, read November 14, 2026.
| Switch | Default | Where | What it covers | What it doesn't cover |
|---|---|---|---|---|
| Data Sharing box | On unless you opt out (X) | X: Privacy & safety, Data sharing and personalization, Grok & Third-party Collaborators (X) | Your public X data and Grok-on-X chats, for xAI and third-party collaborators | Rated chats; a deployed model, which can keep "learning as a result of its normal use"; Grok.com and app chats; X's own license and training |
| Grok Personalization box | Not stated | Same X screen (X) | How X data personalizes Grok for you | Training |
| Protect your posts | Off; posts are public by default (X) | X: Privacy & safety, Audience and tagging | Your posts, from Grok training and from Grok answers (X) | Grok chats; followers' screenshots; authorized apps; copies made while public |
| Grok.com and app: Improve the Model | Training allowed until you turn it off (FAQ) | Grok.com: Settings, Data. App: Settings, Data Controls | Your new Grok.com and app conversations | Past conversations; a rated chat, which "may be used" for training after you opt out; Private Chat is per chat only; no opt-out when logged out in some regions outside the EU and UK (FAQ) |
| Grok Bot in X Chat | Same as the Data Sharing box (X) | X: Privacy & safety, Data sharing and personalization, Grok & Third-party Collaborators | Your chats with Grok Bot | X says Grok Bot "is governed by the Cursor Privacy Policy" |
| Objection in Europe | Not applicable | Your settings, or the privacy portal (addendum) | Training based on SpaceXSI's legitimate interests | Training already done |
For the Grok.com and app settings in detail, including Private Chat and the feedback exception, see is Grok private.
Does making my X account private stop Grok training?
For your posts, yes. X's help page says protecting them prevents them "from being used to train Grok and xAI's underlying AI models or to be surfaced in response to user queries" (X). Turn it on under Settings and privacy, Privacy & safety, Audience and tagging, Protect your posts.
It leaves three things open, according to X's page on protected posts, read from an archived copy dated September 19, 2026:
- Your followers. They can still "capture images of your posts and share them."
- Apps you've authorized. "That third-party application may be able to see your protected posts."
- Copies made while you were public. X "does not have the ability to remove content on websites other than X.com," and its privacy policy says search engines "may retain copies of your posts longer."
Protecting your posts does nothing for your Grok chats, which still need the Data Sharing box and Improve the Model.
Does opting out stop X itself from training on your posts?
No. The Data Sharing box is about Grok and X's collaborators, and X keeps its own rights. X's terms of service, effective October 9, 2026, give X a license to your content that lets it "improve the Services, including, for example, for use with and training of our machine learning and artificial intelligence models." X's privacy policy, effective January 15, 2026, says X may use what it collects "to help train our machine learning or artificial intelligence models," and that it makes public posts available through its APIs "in limited quantities for free" while charging "licensing fees for large-scale access." Neither the terms nor the help page names a switch for X's own training, and none of the switches above stops a crawler or another company from reading a public post.
Opting out also doesn't take back what a model already learned. SpaceXSI's FAQ limits its switch to "your new conversations," and X's page doesn't say its opt-out reaches data already used. We explain why trained models can't easily forget in what is machine unlearning.
What about Grok Bot in X Chat?
The Data Sharing box covers it: X's help page gives the same path for Grok Bot and says "opting out means your data will not be used for training" (X). If you add @Grok to an ordinary X Chat, X says X and SpaceXSI "get access to all the messages sent in that conversation after Grok is added, up until the moment Grok is removed" (X). Grok Bot's own privacy policy and how long it keeps data are covered in Morse vs Grok.
Are the Grok training rules different in the EU and UK?
In four ways, on the pages we read:
- SpaceXSI names X posts as training data. Its Europe privacy addendum, effective August 24, 2026, lists "X Public posts for over 18 year olds" among the data it uses "to train and improve our models," on the basis of its "legitimate interests."
- Logged-out use. The FAQ's warning that some logged-out users can't opt out applies only outside the EU and UK.
- X's terms. The version for the EU, EFTA States, and the UK adds that "if you have chosen via our features to limit the distribution of your Content to a restricted community, we will respect that choice."
- The regulator. On September 4, 2024, Ireland's Data Protection Commission ended its High Court case after X agreed to keep its undertaking to suspend Grok training on EU/EEA users' public posts "on a permanent basis" (DPC). Engadget reported that as X "will permanently avoid training" Grok on those posts, but the DPC's August 8, 2024 statement limits the undertaking to posts processed between May 7 and August 1, 2024, so people in Europe should still clear the Data Sharing box.
People in the EU and UK can also send a formal objection under Article 21 of the GDPR, as our guide to objecting to AI training under the GDPR shows. This isn't legal advice.
Do Grok.com, the Grok app, and Private Chat need their own settings?
Yes. SpaceXSI's FAQ sends Grok-on-X users to X for that opt-out, and X's box doesn't reach Improve the Model, so turn off both; Private Chat keeps one conversation out of training at a time. We go through both, with SpaceXSI's caveats, in is Grok private.
Does deleting my Grok history take it out of training?
Neither company says it does. Guides such as Windows Central's, Tom's Guide's, and BGR's list X's Delete Conversation History link alongside the opt-out, but X's help page says only that deleted conversations "are removed from our systems within 30 days, unless we have to keep them for security or legal reasons" (X). The steps for each surface are in how to delete Grok history.
How Morse keeps your chats out of Grok's training
Each switch above asks SpaceXSI or X to stop using words it already holds. Both keep the chat or the post, train on it by default, and leave it to you to find each box on each surface, skip the thumbs, and accept that past training stays. Morse doesn't train AI on your content, sell your data, or show ads, so there's no Morse switch to find (Morse privacy policy). Morse can't change what you've posted on X, but your chats on Morse aren't tied to your X or Grok account.
- Private, to keep asking Grok under a zero data retention agreement. It's the mode a new chat starts in, and Grok runs in it. Our proxy removes your account, email, and IP address and sends the message under Morse's own account, so no X login or Grok.com account is attached. Morse has zero data retention agreements with its inference, search, and other providers, so they've agreed to keep nothing. Where a provider confirms zero retention on each reply, Morse checks it and refuses a reply that doesn't; otherwise Morse relies on the provider's commitment. Grok 4.7 comes with Pro, at $19 a month.
- E2EE, for questions you'd tell no AI company. Pick an open-weight model like GLM, Kimi, or DeepSeek. Your browser checks the enclave's hardware attestation 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, and the guarantee rests on the enclave hardware and attestation working as designed. E2EE has no web search or image generation.
- Anonymous, for GPT, Claude, and Gemini without your name on it. 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.
In every mode, your chat history is stored in your browser, encrypted, and our servers don't keep messages or log their content. There's no copy on our servers waiting out a 30-day deletion queue. Because we don't store chats, we usually have very little to give authorities. Clearing your browser or deleting your account removes your chats for good, and Morse can't recover them, so keep a copy of anything you need.
Send a first message from the Morse home page with no account, or sign up free and keep your chats on your own device.
Common questions
Does turning off Grok training stop Grok from reading my public posts?
No. The training box decides what goes into future models. X says protecting your posts is what keeps them from being "surfaced in response to user queries."
I use Grok on X and on Grok.com. Do I need to opt out twice?
Yes, two switches: the Data Sharing box on X and Improve the Model on Grok.com and in the app. Our guide to whether Grok is private covers both.
If I protect my posts now, does Grok forget the ones it already trained on?
None of the pages we read on November 14, 2026 says so. X's help page says protecting your posts prevents them from being used to train Grok, without saying it removes posts already used, and copies made while they were public can stay with search engines and other sites.
Does Grok Bot in X Chat follow the Data Sharing box?
Yes. X's help page gives the same path for Grok Bot and says "opting out means your data will not be used for training."