Grok has collected more privacy and safety headlines than almost any other major AI assistant. There have been regulator investigations in the UK and Ireland, a Dutch court order, thousands of shared chats showing up in search results, and a US class action over data sharing. Reporters have also shown that Grok will return a stranger's home address when asked. This guide separates what is documented from the noise, so you can decide how much to trust it.
Does Grok keep my data private?
It depends on which data you mean. This is the most confused part of how Grok works.
Your direct conversations with Grok can be used to train xAI's models, and most coverage we found describes this as on by default, so you would have to opt out. Published guides do not fully agree on the default, and xAI's own policy pages were not accessible to us when we checked, so confirm the current state in your own account before relying on either version. The setting is called "Improve the Model" in the Grok app (Settings, then Data Controls) and on grok.com. On X, the equivalent control sits under Privacy and safety, then "Grok & Third-party Collaborators." Turning it off applies to future conversations only. It does not undo training that already happened.
xAI's policy also lets a limited number of authorized staff review conversations, including for safety and abuse investigation. So "private" never meant "nobody at xAI can see it."
Your public posts on X are a separate matter. If your account is public, Grok can draw on your posts by default. The same X privacy setting controls this, and the opt-out is a toggle you have to switch off yourself.
Private Chat is Grok's temporary mode. These conversations do not appear in your history. xAI's policy says they are deleted from its systems within 30 days, unless they must be kept longer for legal or safety reasons. That is not the same as "never stored."
Is Grok more private than ChatGPT or Claude?
On training defaults, the differences are smaller than most people assume. Consumer ChatGPT also uses conversations to improve its models unless you turn that off. Since late September 2025, Anthropic's consumer Claude plans work the same way: you choose whether chats are used for training, and keeping that on extends retention to five years.
Business plans are different. ChatGPT Enterprise and Claude's work, education, and API products are handled under separate terms, and their data is not used for training by default.
Two real differences do exist:
- A social network layer. Grok sits on top of X, so your public posts are a data source. ChatGPT and Claude have no equivalent.
- Personal lookups. In a December 2025 test by Futurism, ChatGPT, Gemini, and Claude declined simple address requests that Grok often answered (details below).
If you want training controls on ChatGPT, see our guide to the GPT-5 training data opt-out.
Can people see what I do with Grok?
Your one-on-one conversations are not visible to other users by default. Two incidents are still worth knowing about.
The share-link incident. In August 2025, Forbes reported that Grok's "share" button created a public page for each conversation, and that search engines indexed those pages. Forbes counted more than 370,000 indexed conversations. Some contained personal details, and at least one included a password. Users were not warned that sharing meant publishing. A link that anyone can open is not private, even if you never post it anywhere.
The sexual imagery investigations. Grok's image tools were used to create non-consensual sexual imagery of real people, including children. Regulators and courts responded in this order:
- January 2026: Image generation and editing on X was limited to paying subscribers after the backlash. Press reports said misuse continued despite the paywall.
- February 3, 2026: The UK Information Commissioner's Office (ICO) announced formal investigations into X Internet Unlimited Company and X.AI LLC over Grok's potential to produce harmful sexualised images and video.
- February 17, 2026: Ireland's Data Protection Commission (DPC), the lead EU privacy regulator for X, opened its own GDPR inquiry into Grok.
- March 26, 2026: The Amsterdam District Court, in a case brought by the nonprofit Offlimits, ordered X.AI, X Corp, and X Internet Unlimited Company to stop Grok from generating and distributing non-consensual nude imagery and child sexual abuse material in the Netherlands. The penalty is €100,000 per day, capped at €10 million per entity.
These are open investigations, not findings of wrongdoing. The UK investigation is also a data protection matter. Regulators like the ICO look at how personal data is processed, a topic we cover in our UK GDPR guide.
None of this means a stranger can browse your private chat history. It is a good reason not to assume anything you share stays as private as you expect.
Is it true Grok will give out someone's home address?
Yes, in at least one well-documented test. In December 2025, Futurism entered the names of 33 non-public individuals into the free web version of Grok with simple prompts like "[name] address."
- 10 responses returned an accurate, current home address.
- 7 returned an out-of-date address that had once been correct.
- 4 returned a workplace address.
- 12 returned addresses for the wrong people with similar names.
- Only 1 was flatly declined.
Grok also volunteered phone numbers, emails, and details about family members, including children. Futurism reported that ChatGPT, Gemini, and Claude declined comparable requests. xAI did not respond to its request for comment. Other outlets, including Bitdefender and Business Today, reported on the findings, but Futurism ran the test itself.
So the accurate framing is "often, but not always." It is a small test, and it does not prove Grok is a reliable people-search tool. It does show the risk is real.
The standard advice applies. Reduce what is publicly discoverable about you and your family. Data brokers are a major source of that information, and our data broker registration guide explains how US laws treat them.
What's the Grok data-sharing lawsuit about?
The case is Skaggs v. X.AI, LLC (No. 26-cv-04550-BLF) in the US District Court for the Northern District of California. Austin Skaggs filed it as a proposed class action on May 14, 2026.
The complaint alleges that xAI put tracking technology from Google, Meta, and TikTok on Grok.com. It says this sent details of users' queries to those companies, including prompts about finances, health, and legal matters. It also alleges this happened even when users declined cookies.
In August 2026, the judge denied xAI's request to move the case to Texas, finding the terms of service were not displayed conspicuously enough to bind the plaintiff. That ruling is about where the case is heard. It says nothing about whether the allegations are true. The case is ongoing, and the claims are unproven.
Independent researchers at IMDEA Networks Institute also reported in May 2026 that several major AI assistants, not only Grok, send some data to third-party analytics and advertising services. This is not unique to xAI.
Should I use Grok for anything sensitive?
Treat Grok as you would any AI assistant that is still earning public trust. It is fine for low-stakes questions with no personal details.
It is a poor fit for medical, legal, or financial questions, or for anything involving real people you would not want reviewed by staff, used for training, or indexed by a search engine. If you work in an organization, our guide to AI governance covers how to set rules for tools like this.
How to reduce your exposure if you use Grok
- Turn off "Improve the Model" in the Grok app or on grok.com, and turn off the matching Grok setting in X's privacy settings.
- Use Private Chat for anything you would rather not keep in your history.
- Avoid conversation-sharing links for anything you would not want searchable.
- Do not put personal details, passwords, or other people's information in prompts.
- Review your public X profile for details like your address, workplace, and family names.
FAQ
Does Grok train on my conversations?
Most coverage says yes by default: Grok conversations can be used to improve xAI's models unless you turn off "Improve the Model" in Grok's settings, and the opt-out applies going forward only. Some guides describe the default differently, so check the setting in your own account. Your public X posts are also available to Grok by default, and X's privacy settings control that.
Is Grok's data shared with other companies?
A proposed class action, Skaggs v. X.AI, alleges that Grok.com sent conversation details to Google, Meta, and TikTok through tracking technology. The allegations are unproven and the case is ongoing. Check xAI's current privacy policy for its own statement on selling data and advertising use.
Can Grok see private information about me?
It can surface publicly available information about you, sometimes accurately. In one 2025 test, Grok returned a correct current home address for 10 of 33 non-public people. Whether it can reach non-public data is not something xAI has documented in detail, so assume anything online about you could appear.
Is there an incognito mode for Grok?
Yes. Grok has a Private Chat mode that keeps conversations out of your history. According to xAI's policy, those chats are deleted within 30 days unless they must be kept for legal or safety reasons.
If your business is evaluating which AI tools employees can safely use with customer or company data, the consent and data-handling questions here are the ones we help organizations work through. Start with our AI Governance guide to assess tools like this before they end up on an employee's laptop.
