AI privacy and safety: share less, get the same help

Whenever you chat with an AI, your words leave your device and are processed on someone else's computers. That is true of every AI chat service, ours included. It is not a reason to avoid AI, but it is a reason to choose what you send. Below you will find the route a message takes on grok.online, what is better kept out of any chat, how to remove personal details without making your question useless, and how to use AI answers on health, money and legal matters without getting hurt.

What happens to a message you send here

1. Your browser sends it to our server. Together with the earlier messages of the same conversation, so Grok has context.
2. Our server passes it to our AI gateway, which forwards it to OpenRouter, a cloud AI service, and from there to xAI's Grok model, which writes the answer.
3. The answer streams back the same way and appears in your chat.

Neither your messages nor Grok's replies are saved as text on our side. Our gateway keeps technical records, such as which model was used, how large the request was, the time and a hashed version of your IP address, which we use for daily limits and to stop abuse. OpenRouter and xAI process the request under their own terms. Images you attach travel the same route and are used only for that answer.

Your chat history is stored only in your browser on your device, not on our server and not in an account. The exception is a conversation you choose to share: a share link stores a read-only copy on our server until it is removed. The full details are in our privacy policy.

What never to paste into any AI chat

Some information is risky to send to any online service, whoever runs it. Keep these out:

• Passwords, PINs, one-time codes and recovery phrases. No genuine task needs them.
• Full card numbers, bank details and ID or passport numbers.
• API keys, access tokens and private keys hidden inside code or configuration files.
• Other people's private information: a friend's medical history, a client's contract, a colleague's performance review.
• Confidential work material that your employer or a contract does not allow you to share with outside services.
• Your full identity in one place: name, address, birth date and phone number together.

Remove personal details and keep the question useful

You can almost always get the same help with fewer details. Replace what identifies a person with a label that keeps its role.

Risky:
My neighbour Tom Becker at 14 Elm Road keeps parking across my
driveway at number 16. Write him a letter.

Safer, same result:
My neighbour keeps parking across my driveway. We have spoken once,
politely, and nothing changed. Write a short, firm but friendly
letter asking him to stop. Leave [NAME] and [ADDRESS] as blanks.

Grok does not need the real name to write a good letter; you fill the blanks in afterwards. The same works for work documents (use [CLIENT], [PROJECT], [AMOUNT]), for medical questions (age range and symptoms, not your name and hospital number) and for code (YOUR_API_KEY instead of the real key).

When you paste a long document, skim it first for signatures, phone numbers and e-mail footers. They are easy to forget and rarely needed for a summary.

Photos and screenshots

Images often carry more than you notice. Before attaching one:

• crop to the part that matters, such as the error message and not the whole desktop;
• cover names, addresses, faces of other people, number plates and barcodes;
• check screenshots for open tabs, notifications and chat previews in the corners;
• for documents, photograph only the section you are asking about.

Our chat uses an image only to answer the message it came with, but cropping is a good habit everywhere.

Share links, shared devices and exports

A share link creates a public, read-only copy of a conversation. Anyone with the link can open it, and it can be passed on. Read the whole conversation before you share it, and if you change your mind, ask for removal through the contact form. Shared pages are kept out of search engines.

Because history lives in your browser, anyone who uses the same browser profile can read it. On a shared or public computer, delete your chats when you finish, or use a private window, which forgets them when closed. Exported history files are plain files on your device; store them as carefully as any other personal document. If you want help drafting a message about something sensitive, the email writer works fine with placeholders instead of real names.

Health, money and legal questions

Grok can explain terms, summarise options and help you prepare questions. It cannot examine you, see your full situation or take responsibility for advice. In our chat it also cannot check current rules or prices, because it has no internet access.

Weak:
I have chest pain, what should I take?

Better:
Explain in general terms what kinds of questions a doctor usually
asks about chest pain, so I can describe my symptoms clearly at my
appointment. I understand this is not medical advice.

Anything urgent, such as chest pain, breathing trouble or thoughts of self-harm, needs emergency services or a professional, not a chatbot. For money and legal matters, use Grok to understand the vocabulary and list questions, then check with a qualified adviser or an official source before you act.

Scams and manipulation

AI makes it cheap to write convincing scam messages, fake reviews and impersonations. A few habits help:

• Do not trust a message just because it is well written. Check the sender through a channel you already know.
• Paste a suspicious message into Grok and ask: "What signs of a scam does this message show?" It is a useful second opinion, not a verdict.
• Be wary of text you copy from unknown websites into an AI chat; it can contain hidden instructions meant to steer the answer.
• Remember that the AI can be confidently wrong. See our guide on fact-checking AI answers.

FAQ

Do you read my conversations?

No. Message and answer text is not kept on our server, so there is nothing for anyone here to go back and read. Shared conversations are the exception, because you published them with a link.

Is my history synced to the cloud?

No. It stays in your browser. To move it to another device, export it to a file and import it there. Clearing your browser data deletes it for good.

Does xAI see my messages?

Yes, in the sense that Grok has to process a message to answer it: the request reaches xAI through OpenRouter. Our site is independent and is not run by xAI or X. See our privacy policy and what is Grok for background.

Can I use the chat without giving any personal data?

Yes. No account, name or e-mail is needed. Avoid typing personal details in your messages and you can use every feature, from redacting a text to the AI tools.

Is it safe for children?

The site is not intended for children under 13. Older students should follow their school's rules and use the chat with the same care as any other online service.