Skip to content
SafetyScorer

Is Grok safe?

Grok

Grok's app collects less than most AI apps and lets you opt out of training, but its safety record is the weakest of the big chatbots. In early 2026 it was used to make sexualised images of real people, including children, and regulators in the EU, Ireland and the UK opened investigations that are still running.

Medium confidence#48 of 61 appsUpdated
Spot a mistake?
Company
xAI (now SpaceXAI LLC); App Store seller X Corp.
App Store label
No tracking; contact info, identifiers and diagnostics linked to you
Training on your chats
On by default; "Improve the model" toggle or Private Chat
Deleted chats
Removed within 30 days, unless kept for safety or legal reasons
Age rating
16+ (App Store); 13+ with parental permission in xAI's terms
Shared chats
Over 370,000 indexed by Google (August 2025)
Open investigations
European Commission, Irish DPC, UK ICO and Ofcom (2026)

How the score breaks down

The overall score is the weighted average of these factors. Scoring rules

  • Privacy30% of score

    45 out of 100, Weak

    Apple's label lists only contact info, identifiers and diagnostics, with no tracking. But xAI trains on your prompts and uploads by default, staff may review chats, and Google Play lists photos and app interactions shared with other companies.

  • Security30% of score

    50 out of 100, Weak

    Encrypted in transit, not end to end. In August 2025 Forbes found more than 370,000 shared Grok chats indexed by Google, some with names, medical questions and a password, because the share button published them without a warning.

  • Transparency and control15% of score

    65 out of 100, Fair

    An "Improve the model" switch, Private Chat (deleted within 30 days), and in-app download and deletion of chats and account. Training is on by default, and logged-out users outside the EU and UK cannot opt out.

  • Scams and harmful content15% of score

    20 out of 100, Poor

    In early 2026 Grok was used to make sexualised images of real women and children from ordinary photos. The European Commission said those risks "seem to have materialised". xAI says some modes can produce sexual content; the App Store rates it 16+.

  • Regulatory record10% of score

    35 out of 100, Poor

    No fine yet. Ireland's DPC took xAI's training on EU X posts to the High Court in 2024 and opened a GDPRThe EU's General Data Protection Regulation, which sets rules on how organisations collect and use personal data and allows large fines.More in the glossary inquiry into the Grok images in February 2026. The EU (DSA), UK ICO and Ofcom are investigating, and California's attorney general sent a cease-and-desist letter.

Is Grok safe?

Grok is a capable chatbot with a weaker safety record than its rivals. The app itself collects relatively little on paper, and you can stop xAI training on your chats. The problem is what Grok has been willing to produce. At the turn of 2026 it was used, mostly through the @Grok account on X, to turn ordinary photos of real women and children into sexualised images. Regulators in Brussels, Dublin and London opened investigations, and none had closed by October 2026. We score it 45/100, below ChatGPT and above X, the platform it is built into.

The deepfake scandal and who is investigating

At the turn of 2026, users discovered that they could reply to a photo on X and ask Grok to "undress" the person in it. Grok complied, publicly, thousands of times. Some of the people were children. The response from regulators:

DateRegulatorAction
12 January 2026Ofcom (UK)Formal Online Safety Act investigation of X
16 January 2026California attorney generalCease-and-desist letter to xAI, according to Reuters
26 January 2026European CommissionDSA investigation into whether X assessed Grok's risks before launch
3 February 2026ICO (UK)Data protection investigations into X and into X.AI LLC, the company behind the Grok app
17 February 2026Data Protection Commission (Ireland)Large-scale GDPRThe EU's General Data Protection Regulation, which sets rules on how organisations collect and use personal data and allows large fines.More in the glossary inquiry into X over the images

The Commission said the risks "seem to have materialised, exposing citizens in the EU to serious harm". Malaysia and Indonesia blocked Grok for a time. X and xAI restricted the feature, but the investigations stayed open, and an EU ban on AI systems that generate non-consensual intimate images applies from 2 December 2026.

The standalone app has its own adult side. xAI's privacy policy warns that, depending on the mode a user picks, Grok may produce "suggestive dialogue, coarse language, crude humor, sexual situations, or violence". Apple rates the app 16+.

What xAI does with your chats

On paper the Grok app is restrained. Apple's privacy label lists no tracking and only three kinds of data linked to you: contact info, identifiers and diagnostics. That is a shorter list than ChatGPT's or Meta AI's. Google Play's data safety form is fuller: it lists approximate location, photos, files, search history and other content you create, and says app interactions, photos and device IDs may be shared with other companies for analytics and security.

The bigger issue is training. xAI's FAQ says it may use your prompts, searches, uploads and Grok's answers to train its models, and this is on by default. xAI also trains on public X posts. In 2024 Ireland's Data Protection Commission went to the High Court because xAI had started training Grok on EU users' posts before giving them a way to object; the case ended in September 2024 when X agreed to keep to its undertaking to stop. A small number of xAI staff can review conversations for model improvement and misuse checks.

xAI says it does not sell your data or share it for advertising. Its privacy policy does say its website cookies may be used for "targeted advertising".

The shared-chat leak

In August 2025 Forbes found that Google had indexed more than 370,000 Grok conversations. Every time someone pressed "share", Grok created a public page that search engines could crawl, with no warning. Some pages included names, medical and psychological questions, uploaded spreadsheets and at least one password. It was not a hack, but it shows how easily a private chat can become public. ChatGPT had a similar problem in 2025 and removed the option.

Grok chats are encrypted in transit but not end to end, so xAI can read them.

How to use Grok more safely

  1. In the app, go to Settings, Data Controls and turn off "Improve the model". On grok.com the same switch is under Settings, Data.
  2. Use Private Chat for anything sensitive. It is not saved to your history, not used for training and deleted within 30 days.
  3. Never press share on a conversation that contains personal details.
  4. If you use Grok inside X, change the Grok training setting in X's privacy settings too; the app setting does not cover X.
  5. Do not upload photos of other people, and never photos of children.
  6. Keep it off a child's phone. The 16+ rating is there for a reason.

How we scored it

Scams and harmful content (20) drags the score down: Grok itself generated sexualised images of real people, which regulators in three jurisdictions are now investigating. Regulatory record (35) reflects those open cases and the 2024 High Court action, though no fine has been issued. Transparency and control (65) is Grok's best factor, thanks to a clear training switch, Private Chat and 30-day deletion.

Questions people ask

Does Grok use my chats to train its AI?

Yes by default. In the Grok app go to Settings, Data Controls and switch off "Improve the model", or use Private Chat (the ghost icon), which is not used for training. If you use Grok inside X, the training setting is in X's own privacy settings. Feedback you send can still be used for training.

Can Grok make fake nude images of people?

It was used to do exactly that at the turn of 2026, mostly through the @Grok account on X, including images of children. X and xAI restricted the feature, but the EU, Ireland's DPC, the UK's ICO and Ofcom opened investigations, and none had reached a decision by October 2026. Sharing intimate images of a real person without consent is a crime in the UK and many other countries.

Is Grok safe for kids?

No. The App Store rates it 16+, and xAI's own policy warns that, depending on the mode, it may produce suggestive dialogue, sexual situations or violence. Teenagers aged 13 to 17 need a parent's permission under xAI's terms. We would not give it to a child.

Are my Grok conversations private?

They are not End-to-end encryptionEncryption in which only the sender and the recipient can read a message. The company running the service cannot see its content.More in the glossary, and xAI says authorised staff may review them. Shared chats can be public. In August 2025 Forbes found more than 370,000 shared Grok conversations in Google results. Do not press share on anything personal.

Is the Grok app the same as Grok on X?

Same model, different rules. The standalone app and grok.com are covered by xAI's privacy policy. Grok inside X is covered by X's privacy policy, and its training and deletion settings live in X.

Sources

  1. 1.Grok AI App Store listing and App Privacy label (opens in a new tab) · Apple App Store, 2026
  2. 2.Grok – Data safety (opens in a new tab) · Google Play, 2026
  3. 3.SpaceXAI Privacy Policy (effective 24 August 2026) (opens in a new tab) · SpaceXAI, 2026
  4. 4.Consumer FAQs (opens in a new tab) · SpaceXAI, 2026
  5. 5.Data Protection Commission opens investigation into X (XIUC) (opens in a new tab) · Data Protection Commission (Ireland), 2026
  6. 6.ICO announces investigation into Grok (opens in a new tab) · Information Commissioner's Office, 2026
  7. 7.Commission investigates Grok and X's recommender systems under the Digital Services Act (opens in a new tab) · European Commission, 2026
  8. 8.Data Protection Commission welcomes conclusion of proceedings relating to X's AI tool 'Grok' (opens in a new tab) · Data Protection Commission (Ireland), 2024
  9. 9.Elon Musk's xAI published hundreds of thousands of Grok chatbot conversations (opens in a new tab) · Forbes Australia, 2025
Spot a mistake?