Are your Meta AI chats public? Only the ones you shared or posted. A normal back and forth with Meta AI is not broadcast to strangers by default, but the app ran a public Discover feed where any conversation you tapped Share on could show up for the world to read, and some of those shared chats carried home addresses, medical questions, and other sensitive details. So the honest answer to is Meta AI private is: it is private until you press a button that makes a chat public, and that button was easy to press by accident.
This post explains plainly what happened, why it was not a hack, how to check whether any of your Meta AI conversations are exposed, how to delete them or lock them down, and what changed in July 2026 when Meta added a private path for talking to its assistant. No fearmongering, just the steps and the honest framing.
What actually happened: public by design, not a breach
No one broke into Meta AI. The exposed conversations reached the public Discover feed because users pressed a Share button that published their chat, often without realizing what public meant. Security researchers were blunt about this: the chats were public, and it was not a bug. Meta AI would show a preview screen after you tapped Share, and from there a post went live in a feed that anyone could scroll, in some cases without even logging in.
The problem was the design, not a security failure. Reporting described a flawed user experience rather than a breach: minimal warnings, no strong double check, and a share flow that many people read as save or bookmark rather than publish to everyone. The result was that ordinary users posted things they clearly thought were private.
The contents were the alarming part. Scans of the Discover feed surfaced medical questions about post surgery symptoms, legal messages about job terminations and eviction notices, prompts about tax matters, personal home addresses, and intimate concerns about marriages, sometimes tied to real usernames and profile photos. In April 2026, TechCrunch reported on the same pattern and noted that Meta had since removed the Discover feed.
The reaction was loud enough that it moved beyond commentary. Privacy advocates and EU regulators pressed for stricter controls or a temporary suspension of the feed, and the Mozilla Foundation publicly demanded Meta shut down the Discover feed until it stopped turning private AI chats into public posts. Meta responded by adding a pop up warning that any prompt you post is public and visible to everyone, and later by removing the Discover feed altogether.
Why did this hit so hard? People treat chatbots like a private confidant. They ask an assistant the things they would not say out loud, from a health scare to a money worry to a question about a relationship. When the same product that invites those confessions also offers a one tap route to a public feed, the mismatch between what users expect and what the app does is where the damage happens. The technology worked as built. The expectation did not match the design.
The key distinction: a shared Meta AI chat was public because someone posted it, not because someone stole it. That makes it preventable. It also makes a public chatbot feed a poor place to type anything you would not want a stranger to read.
How to check whether your Meta AI chats are exposed
Start by checking whether you ever posted a prompt publicly, because that is the only way your Meta AI chats would be visible to others. Open the Meta AI app or the assistant inside Facebook, Instagram, or WhatsApp, then work through your history and settings. If you never used the Share or Post button, there is nothing in a public feed to worry about.
- Open the Meta AI app, tap your profile icon, then go to Settings, then Data & Privacy, then Manage your information. This is where public prompt controls live.
- Look for any history of posted or shared prompts. Anything you sent to a feed will be listed as public rather than private.
- Remember that Meta AI reached you through several apps. If you chatted with the assistant in Facebook, Instagram, or WhatsApp, check each one, because a prompt you shared in any of them could be tied to that account.
- If you are not sure whether you ever tapped Share, assume you might have and go straight to the lock down steps below. It costs nothing to set everything to private.
Searching your own name inside the Meta AI Discover surface used to be a quick way to spot your posted prompts. Since Meta removed that feed, the reliable check is the Data & Privacy settings screen, which shows and controls what you made public.
How to make your Meta AI chats private and delete them
The fastest fix is one setting: make all your prompts visible to only you. That single toggle pulls your shared prompts out of public view and stops future ones from going to a feed. From there you can delete old conversations and turn off the sharing habit entirely.
- In the Meta AI app, tap your profile icon, then Settings, then Data & Privacy, then Manage your information, then choose Make all your prompts visible to only you.
- Stop using the Share button after a chat unless you genuinely want that exchange posted where anyone can read it.
- To wipe conversation history in a messaging app, type /reset-ai in your Meta AI thread inside Messenger, Instagram, or WhatsApp to clear that chat.
- To limit how your interactions feed model training, open Settings & Privacy, then Privacy Center, then Privacy Topics, then AI at Meta, and submit an objection request.
- Delete individual posted prompts you no longer want visible, then confirm in the same Data & Privacy screen that nothing remains marked public.
Menu names shift between app versions, so if a label reads slightly differently, look for the nearest match under Data & Privacy or Manage your information. The two things that matter are the same everywhere: set prompts to visible to only you, and stop posting chats to any feed.
One more habit worth building: separate the objection request from the sharing setting in your head, because they fix different problems. Making prompts visible to only you controls who else can see a chat. Submitting an objection under AI at Meta controls whether your interactions feed model training. You want both if your conversations touch anything personal, and neither one undoes the other.
The July 2026 change: a private path inside Threads
On July 27, 2026, Meta made talking to its assistant more private by bringing Meta AI into Threads direct messages, so a conversation with the bot stays in your DMs instead of anywhere near a public feed. You can send Meta AI a post, image, link, or video and ask follow up questions, all inside a private message thread rather than a broadcast surface.
This is a real improvement over the old share flow, because the default location of the chat is a private message rather than a feed you might accidentally post to. The reporting framed it directly: AI chats in DMs are more natural and inviting than public feed conversations with a chatbot, which can be spammy and, well, public. Fewer chances to broadcast something by mistake is a genuine step forward.
Private inside an app is not the same as private from the company. A DM with Meta AI is not posted to strangers, but it still runs on Meta's servers under Meta's data policies, and your interactions can still inform the wider system unless you opt out. Private path is about who else can see the chat in the app, not about the assistant forgetting what you typed.
Where a sensitive question is actually safest
If you are deciding where to put something personal, the surface matters more than the wording of any one setting. The table below compares the old public share flow, the newer private assistant path, and a memory app built to be private by architecture, so you can see the trade in exposure at a glance.
| What to consider | Meta AI public share / old Discover feed | Meta AI private DM in Threads (July 2026) | Private by architecture memory app (MemX) |
|---|---|---|---|
| Who can see the chat | Anyone, once you tap Share to post it | Just you and the assistant in the DM | Only you; nothing posts to any public feed |
| Risk of accidental public post | High: a Share button read as save or bookmark | Low: the default surface is a private message | None: there is no feed to publish to |
| Was it a hack | No: public by design, posted by the user | No: a private surface by default | Not applicable: no public sharing surface exists |
| Data isolation and training | Runs on Meta systems; opt out separately | Runs on Meta systems; opt out separately | Per user isolation, encryption at rest, no training on your data |
None of this means Meta AI is useless. For casual questions it is fine, and the Threads DM path removes the most obvious way people got burned. The point is narrower: any assistant that ships with a one tap route to a public feed is the wrong home for a medical worry, a legal document, or an address you would not shout across a room.
The honest takeaway, and where sensitive stuff belongs
The lesson from the Discover feed episode is simple. The danger was never a mysterious breach. It was a normal app where the line between private note and public post was one confusing tap wide, and enough people crossed it that strangers could read their health and money problems. Meta added warnings, removed the feed, and built a private path, which is progress. Still, the safest move for anything sensitive is to keep it out of a product whose business is a social feed.
That is the gap MemX is built for. MemX is a private memory app for the personal things you actually need to find later: documents, photos, voice notes, and messages, searchable by simply asking a question and getting the answer with its source. It is private by architecture, meaning per user isolation, encryption at rest, on device processing where possible, and no training on your data. There is no Discover feed, no Share to everyone button, and nothing that turns a private note into a public post. Snap it now, find it later, without wondering who else can read it.
Rule of thumb: if a tool has a public feed, treat everything you type into it as potentially postable. Put your genuinely private information somewhere that has no way to publish it in the first place.
01Are my Meta AI chats public?
Not by default. A normal chat is not broadcast to strangers. Chats only became public when a user tapped Share or Post, which sent that conversation to the app's public Discover feed. If you never shared a prompt, nothing of yours is in a public feed.
02Was the Meta AI Discover feed exposure a hack?
No. Researchers and reporters described it as public by design, not a breach. The chats appeared because people pressed a Share button, often thinking it saved the chat. It was a confusing user experience, not stolen data, which also makes it preventable.
03How do I make my Meta AI chats private?
In the Meta AI app, open your profile, then Settings, then Data & Privacy, then Manage your information, and choose Make all your prompts visible to only you. Then stop using the Share button. Menu labels vary slightly by version, so look for the nearest match.
04How do I delete my Meta AI conversation history?
In a messaging app, type /reset-ai inside your Meta AI thread in Messenger, Instagram, or WhatsApp to clear that conversation. You can also delete any prompts you posted publicly and confirm in the Data & Privacy screen that none remain marked public.
05Is Meta AI inside Threads DMs actually private?
It is more private than the old share flow because the chat stays in your direct messages instead of a public feed, added on July 27, 2026. It is still not private from Meta itself, since it runs on Meta's systems and can inform training unless you opt out.
