How to remove Meta AI from WhatsApp: you can't, but here is what works in three minutes

The search for how to remove Meta AI from WhatsApp always ends the same way: there is no button. WhatsApp describes Meta AI as an optional service, but optional means you are not obliged to use it, not that you can uninstall it. WhatsApp's own help centre explains how to mute the chat, how to delete conversations, how to ask Meta to delete its copy and how to lock a chat so nobody can call the AI into it. A switch that makes the circle disappear from the chat list is nowhere to be found.

What Meta AI reads, and what it doesn't

This is the part that frightens people most and deserves it least. Personal WhatsApp messages stay end-to-end encrypted, and WhatsApp states that Meta can only read messages that mention @Meta AI or that someone chooses to share with the AI. That is where the family angle sits: sharing includes forwarding. A child who forwards a friend's message to Meta AI to ask what it means has just taken that message out of the encrypted chat and sent it to a company. That is not a security flaw, it is the feature working as designed.

Three minutes on your child's phone

• Mute the contact: open the Meta AI chat, tap the name, Notifications, Mute, Always. The circle stays, but it goes quiet and stops pushing itself to the top.
• Look together at what has already been told to it: inside the Meta AI chat sits the history of conversations. Press and hold one and delete it. To ask Meta to delete its copy too, type /reset-ai in that chat, or /reset-all-ais for every chat with the AI at once.
• Settings, Chats, Private Processing: leave it off. That setting controls message summaries and writing help, it ships switched off, and according to WhatsApp it currently only works in English, Spanish, Portuguese and Indonesian.
• In the family chat, if you are the group admin: Group info, Advanced chat privacy, on. From then on nobody in the group can ask Meta AI for answers, images or summaries of that conversation, and the chat cannot be exported either.
• In your one-to-one chat with your child: Contact info, Advanced chat privacy. In individual chats either participant can switch it on.

In groups, anyone can call it

This is what surprises parents in the class group most. In any group where Meta AI is available, any member can type @Meta AI followed by a question, and that message does go to Meta. On some devices Meta AI can also be added as a group member: in that case, according to WhatsApp, it reads and responds to every new message in the group, and only an admin can remove it. So the advice for the parents' group and for the class chat is the same: whoever runs the group switches on advanced chat privacy, and every so often somebody glances at the member list.

The family chat rule fits in one sentence a ten year old understands first time: Meta AI is not a friend, it is an employee of a company who takes everything you tell it back to the office. Three concrete rules follow from that. What is said in the family chat is not forwarded to Meta AI. In groups, nobody types @Meta AI without asking first. And questions you would not want anyone to read, about health, about friends, about home, are not asked of any AI that comes built into a messaging app.

Remember that Meta AI isn't a human, even though it may chat and respond in ways that seem like a real person.

And if your child has been chatting with it for months

Most likely they have, and they probably don't even count it as using AI, because to them it is just another contact in the list. That is not a catastrophe. Open the history together, without the interrogation voice, and look at what was asked: it is almost always homework, jokes and things that felt embarrassing to ask at home. That last one is the conversation worth having. Then delete what doesn't belong there, mute the contact, and say the family chat rule out loud, once, without the lecture.