Does xAI train on your data?
Grok trains on your conversations by default. The opt-out is a real settings toggle — documented in xAI's consumer FAQ, which the policy points to; the policy itself never states one.
A self-serve opt-out exists
"To develop and improve our Service and to conduct research: For example to develop new product features, to train our models, to identify usage trends, to operate and expand our business activities, to identify new customers, and for data analysis."
- 7 Sept 2026
- 5 Oct 2026
- All accounts (Grok consumer)
- global
- Fetched from the company
- 851A4BCF…0DF2
How we read it: Regraded F to D on 1 September 2026, the day xAI's reply pointed at its Consumer FAQs — a document the privacy policy cites in its opening paragraph for "data controls", and which this entry had not read. That was our miss: the FAQ documents the toggle, and the capture cited predates the grading by a fortnight. Its limits are stated too: opting out binds new conversations only, feedback survives it, and logged-out use in some regions outside the EU and UK has no opt-out at all. The policy was republished on 24 August 2026 under the company's new name, SpaceXAI LLC; the clause is character-identical in the renamed document.
opt out, opt-out, Data Controls, Improve the model, Private Chat, select whether, training
What the document does offer
- Consumer FAQ, wayback capture of 2026-08-15: "On the mobile app, to select or deselect using your content for model training, go to Settings, Data Controls, and select or deselect \"Improve the model.\""
- same FAQ, for the web: "For the Grok.com website, you can go to Settings, Data, and then \"Improve the Model\" to select whether your content is used for model training."
- same FAQ, on scope: "Once you select that your data should not be used for training, your new conversations will not be used to train our models." — forward-looking only, with nothing said about conversations already used
- same FAQ, surviving the opt-out: "Please note that, even if you opt out of model training, when you subsequently decide to voluntarily provide feedback, that feedback may be used for training purposes."
- same FAQ, an alternative the toggle does not require: "When using Private Chat, where available, your content and interactions are not used for model training."
- same FAQ, who has no control: "in some regions (excluding the EU/UK), when you use Grok without logging in, you won't have the option to opt out of model training."
- same FAQ, on human review: "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." — not stated to be reachable by the training toggle
- same FAQ, on business customers: "Similarly, we do not use content from our business and enterprise customers to improve our models."
Why none of it applies: The toggle is real and documented — in the FAQ the policy's first paragraph points at, not in the policy itself. The F rested on "no stated opt-out", and that fails against a document this site should have followed the pointer to on 30 August and did not. The FAQ capture cited is the Internet Archive's copy of 15 August 2026, xAI-branded and pre-dating both the rebrand and this entry: the control was published before we graded.
What xAI (Grok) said when we showed them this
Corrected us“We 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. ... You control whether your data is used for training Grok. ... When using Private Chat, where available, your content and interactions are not used for model training. Similarly, we do not use content from our business and enterprise customers to improve our models.”
What each plan can refuse
The same questions, asked of every plan we could identify. Every cell comes from a quoted clause or a documented search - see the evidence above.
- Self-serve toggle
- Not stated
- 55 / 100
- Not stated
- Yes
How to opt out of xAI training
Plan by plan, from what xAI's own policy says about refusing. Where it publishes the steps, they're listed here with the limits they carry. Where there's nothing to switch off, or nothing graded, this says so instead of inventing a menu path.
All accounts (Grok consumer)
Yes. xAI has a self-serve setting.
- Open Settings, then Data Controls.
- Deselect "Improve the model".
New conversations only. Feedback you give can still be used, and Grok used without logging in has no switch outside the EU and UK. Private Chat (the ghost icon) is the other route xAI names.
- Open Settings, then Data.
- Deselect "Improve the Model".
What this policy has done since we started watching
4 changes detected since 1 Aug 2026. Each one is a diff between two archived copies, and each links to the passage that moved.
Vendor response
Their reply is quoted in full at the top of this page.