Is Grok private? Private Chat, X's settings, and the share-link leak

The Morse teamPublished October 19, 202611 min read

As of October 2026, Grok is private from other people unless you make a share link, but it isn't private from SpaceXAI, the company that runs it. SpaceXAI keeps your Grok chats on its servers, trains on them unless you turn off "Improve the model," lets a limited number of staff review them, and deletes them within 30 days of your request, with exceptions. Grok's Private Chat mode keeps a conversation out of your history and out of training, then deletes it within 30 days.

How private Grok is also depends on where you open it. Grok.com and the Grok apps follow SpaceXAI's own privacy policy, while Grok inside X follows X's. We read both companies' policies, SpaceXAI's FAQ and cookie policy, X's Grok help page, and Grok's plans page on October 19, 2026. SpaceXAI's FAQ does one thing well. It names each setting and says where to find it.

Grok privacy at a glance

Can you use Grok privately with Private Chat?

Partly. Private Chat is Grok's version of an incognito window, and you start it with the ghost-shaped icon at the top right of the screen (Grok FAQ).

• What it does. A Private Chat doesn't appear in your conversation history, and SpaceXAI says it "will be deleted from SpaceXSI systems within 30 days" (Grok FAQ). (SpaceXSI is the short name the policy uses for SpaceXAI LLC.)
• Training. "When using Private Chat, where available, your content and interactions are not used for model training" (Grok FAQ).
• What it doesn't do. It doesn't keep the chat off SpaceXAI's servers. The privacy policy allows a Private Chat to be kept past 30 days if "it is necessary that they be kept longer for legal, compliance, or safety purposes" (privacy policy).
• Staff review. The FAQ's answer on human review makes no exception for Private Chat (Grok FAQ).

Does Grok train on your chats?

Yes, by default. SpaceXAI's FAQ says it "may use your content and interactions with Grok (e.g., prompts, searches, and other materials you submit) along with Grok's responses to train our models" (Grok FAQ).

To turn it off:

• In the Grok app: go to Settings, then Data Controls, and deselect Improve the model.
• On Grok.com: go to Settings, then Data, and deselect Improve the Model.

Four details decide how far that switch reaches, all from the same FAQ:

1. It covers "your new conversations," not ones already used.
2. A thumbs up or down still counts: "even if you opt out of model training, when you subsequently decide to voluntarily provide feedback, that feedback may be used for training purposes."
3. If you use Grok without logging in, "in some regions (excluding the EU/UK)" you "won't have the option to opt out of model training."
4. Business and enterprise customers are excluded: "we do not use content from our business and enterprise customers to improve our models."

A separate switch, Personalize Grok using X, decides whether your X data shapes Grok's answers. It sits next to the training switch in the app's Data Controls and under Data on Grok.com (Grok FAQ).

Is Grok on X private? It follows X's rules, not SpaceXAI's

SpaceXAI's privacy policy says that "use of Grok on the X platform, is governed by the X Privacy Policy and X Terms, not this SpaceXSI Privacy Policy" (privacy policy). SpaceXAI also says it is "a separate company from X Corp." So the Grok tab in X has its own switches, in X's settings.

X's help page says X "may share with xAI your public X data as well as your user interactions, inputs and results with Grok on X to train and fine-tune Grok" (X Help Center). To opt out in X, open Settings, then Privacy & Safety, scroll to Data sharing and personalization, choose Grok & Third-party Collaborators, and under Data Sharing clear the box that allows your data "to be used for training and fine-tuning."

That opt-out has a limit X states itself. When you use an X feature powered by Grok, such as recommendations, "the opt-out does not prevent a deployed model from learning as a result of its normal use" (X Help Center). X's privacy policy, effective January 15, 2026, adds that unless you opt out, third-party collaborators that receive your data may use it "to train their artificial intelligence models, whether generative or otherwise" (X).

One more X rule covers group chats. X says its Chat conversations are end-to-end encrypted, but "if you add @Grok as a member of a Chat conversation, X and SpaceXAI (to provide the feature) will get access to all the messages sent in that conversation after Grok is added, up until the moment Grok is removed" (X Help Center).

Grok.com versus Grok in X, setting by setting

Signing in to Grok.com with your X account connects the two. The privacy policy says you may direct X to share "your Grok on X conversation history" with SpaceXAI, along with your X username and whether you subscribe to X Premium (privacy policy). Grok's paid plans are also bundled with X: as of October 2026, the plans page lists Plus at $8 a month, and Premium, Super, and Ultra at $30, $100, and $200 a month after a discounted first month, each with X benefits such as fewer timeline ads or ad-free scrolling on most surfaces (Grok plans). None of the pages we read gives paying subscribers different data rules. The training exclusion covers business and enterprise customers only.

Can people see your Grok conversations?

Only if you share them, and a share link is a public web page. In August 2025, Forbes and then Fortune reported that "more than 370,000 Grok chats have been indexed by search engines" after users pressed Grok's share button, which "creates a unique URL for the conversation" (Fortune, August 22, 2025). Fortune found "hundreds of sensitive prompts that include medical and psychological questions," business details, and at least one password. It also reported that, unlike OpenAI's, Grok's share function had no disclaimer at the time.

SpaceXAI's FAQ now carries a warning: "if you share the link publicly on a social media platform, it may be subject to indexing by a search engine (e.g., Google) just like any other publicly shared content" (Grok FAQ). You can remove any or all of your links at grok.com/share-links by choosing "Remove" next to each one.

To show someone part of a Grok chat, copy the text instead of sharing the link. A link stays live, and searchable once it's posted anywhere public, until you remove it. Search engines may also keep their own copies of pages that were public, as X's privacy policy notes about deleted posts (X).

Who at SpaceXAI can read your Grok chats?

Machines first, then some people. SpaceXAI says it "may use automated content classifiers and safety tools" to check use against its terms (Grok FAQ). It adds that "a limited number of our authorized personnel may review your conversations with Grok for specific business purposes, including improving model performance, investigating security incidents and potential misuse of our services, and complying with our legal obligations."

The FAQ doesn't say whether turning off "Improve the model" stops review for model performance, and it doesn't say how reviewed chats are stored.

How long does Grok keep deleted chats?

Up to 30 days, with three exceptions that can keep a chat longer:

• De-identified copies. The FAQ says deleted chats and Private Chats are removed within 30 days "unless they have been de-identified or pseudoanonymized and disassociated from your account" (Grok FAQ). The privacy policy says SpaceXAI may use de-identified information to analyze the service, add features, and research, and "will not attempt to reidentify the information, unless required by law" (privacy policy). Neither page gives a time limit for those copies.
• Safety, security, or legal reasons, in the FAQ's words.
• A court order. "If we receive a court order, we would retain your information for longer than our usual retention periods" (privacy policy).

The guides we read in the top search results for "is grok private" give the 30-day figure without these exceptions. A de-identified copy loses its link to your account, but the words you typed stay as you wrote them, so a name or an employer mentioned in a chat stays in that copy. We cover what courts can demand from AI companies in can AI chats be subpoenaed. This isn't legal advice.

Does Grok use your chats for ads?

SpaceXAI's pages don't say chat content picks ads. Its privacy policy does say cookies on its services are used to "deliver relevant content and targeted advertising" (privacy policy). The cookie policy, effective October 12, 2026, lists non-essential advertising cookies from Meta, Google, X, TikTok, LinkedIn, and Microsoft on grok.com, used "to track the effectiveness of specific ads on third-party properties through to SpaceXSI account sign ups and subscriptions" (cookie policy). To refuse them on grok.com, open Settings, then Data Controls, and choose Manage under Cookie Settings.

Grok's privacy policy in 2026: a short timeline

• April 4, 2026. The policy in force names the company X.AI LLC, or "xAI" (April 2026 version).
• August 24, 2026. A new version renames it SpaceXAI LLC, shortened to "SpaceXSI." Apart from the name, we found no change to the wording (August 2026 version).
• October 11, 2026. The current version drops a bold line from the August text, "We ask that you do NOT include personal information in your prompts and inputs into our Service," and adds payment wording for purchases made through the service (current policy). Its contact section still lists "x.AI LLC" for privacy requests.

The FAQ still carries a date of May 12, 2025, though it now uses the new name (Grok FAQ), so its date doesn't tell you when it last changed.

How Morse keeps your Grok chats private

Every gap above starts with two facts. Your Grok chats sit on SpaceXAI's servers, and they're tied to an account, often your X account, under settings split across two companies' policies. Morse takes both away. It's a web app that offers over 30 models, open and closed, with Grok among them: your history stays on your device, and Grok gets your message under Morse's account, through a host that agreed to keep nothing.

• Private, to use Grok through a service that agreed to keep nothing. It's the mode a new chat starts in, and Grok runs in it under Morse's zero data retention agreements with its providers. 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. Our proxy also removes your account, email, and IP address and sends the message under Morse's own account, so there's no X login, no "Improve the model" switch, and no X setting to hunt for.
• E2EE, for words no AI company should read. Grok doesn't run in E2EE, so for those chats pick an open-weight model like GLM, Kimi, or DeepSeek. Your browser checks the secure enclave's hardware attestation before anything leaves your device, then encrypts the message so only that verified enclave can read it. Morse's servers relay ciphertext they can't read. The guarantee rests on the enclave hardware and attestation working as designed. E2EE chats come with Pro, at $19 a month; see how end-to-end encrypted AI chat works.
• 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 all three modes, your chat history is stored in your browser, encrypted, not on Morse's servers. It locks when you sign out, and opening it takes a key that never leaves your device and a second one our server hands over when you sign in. Morse doesn't train AI on your content, sell your data, or show ads, and it runs no advertising trackers or third-party analytics (Morse privacy policy). Because we don't store chats, we usually have very little to give authorities. Deleting a chat or clearing your browser removes it for good, and Morse can't recover it either, so keep a copy of anything you need.

So is Grok private enough for you?

• Other people: yes, if you never press Share. If you have, check grok.com/share-links and remove what you don't want public.
• Model training: turn off "Improve the model" in Grok and the Data Sharing box in X, and skip the thumbs up and down. On Morse, there's no switch, because Morse doesn't train on your chats.
• SpaceXAI's staff and servers: Private Chat shortens how long a chat is kept, but it's still on SpaceXAI's systems for up to 30 days. Morse's Private mode sends Grok your message through a host that agreed to keep nothing, under Morse's account instead of yours.
• Ad trackers: refuse the non-essential cookies on grok.com. Morse doesn't run any.

For how Grok compares with ChatGPT, Claude, Gemini, and Meta AI, see are AI chats private, and for more habits that work in any app, read how to use AI privately.

Send a first message from Morse's home page as a guest, with no account, or sign up free to keep your chats on your own device.