A sensitive conversation about your wife, or a call for help about your daughter, or a strange pocket dial that recorded a conversation with your friend – is unlikely to end in a public feed of Meta's AI app.
Mark Zuckerberg's company is offering users new pop-up warnings before sharing anything with the Meta AI “discovery” feed.
This comes after Business Insider reported last week that its feed was full of personal information that appears to be incorrectly shared. Posts showing parents seeking advice from AI about their whimsical daughter, pocket dial audio where someone chats with a colleague about working hours, and texts from employees seeking legal advice about their employer.
Meta did not respond to Monday's request for comment when asked about the new pop-up.
After last week's Bi story, the BBC, Washington Post, Wired, TechCrunch and other media outlets ran their own stories about the obviously accidental oversharing.
A new warning message that appears on Monday will appear after you press “Share” in the Meta AI app chat. The message says, “The prompt to post will be published and displayed to everyone. Meta can make suggestions on other meta apps. Avoid sharing personal or sensitive information.”
After that warning, the Post to feed button will be disabled until you tap again in the center of the screen. only after that Can I post to the public discovery feed?
Meta AI feeds had some creepy personal information
Old meta AI feeds were littered with these personal messages. Meta AI / Business Insider / Screenshot
Meta AI, a standalone AI chatbot, has a “discovery” feed to share images you create with AI and post chats with your bot. These chats are not public by default. I had to click “Share” and then “Post” to send the chat to the Discover feed.
Obviously, some users didn't understand what they were doing with their previous systems. (One guy I spoke to last week said he had no intention of posting a conversation with the chatbot for car repairs.)
When the app first launched in late April, I wrote about the issue of accidental public posts on meta AI. I checked back last week and realized that the problem seemed to be getting worse. Now there were more posts that seemed to be mistakenly clearly shared. Some of these included intimate personal issues, such as asking the judge to draft a letter about the child's custody, medical issues, and difficulties in the relationship.
People in X also found examples of people posting tax information. Or, like the person who asked what would happen if they placed a deep heat cream on the private part, but asked the bot to keep the conversation private, it seems like he doesn't even understand what he's using. Hmm.
Meta AI's new pop-up warns people before posting
There is a new warning message at the bottom of the screen before pressing “Post Feed”. Meta AI / Business Insider / Screenshot
Of course, it is entirely possible that some of these embarrassing pure posts are secretly intentional. (People who asked Meta AI to do various images of celebrities doing strange things, including Ana de Armas, who put Ana de Armas in bed, might have wanted people to see his stuff.)
Another change to the Meta AI discovery feed I noticed on Monday: It's now almost exclusively an image. In the days when I checked in the app since the start of last week, and in late April, my feed had a massive amount of filling of text and audio-based posts, like a seemingly accidental pocket dial.
Also, Meta did not respond to BI if the apparent tweaks in the discovery feed were intentional (where something other than the image is buried or disappeared) or if the change was a potential natural result of a new extra tap needed to share something with the public feed. Or if it's something completely different.
I'll update this story if Meta comes back to me, but for now I'm happy that I don't have to close my eyes, cling to, flip or fish (similarly) as I scroll through Meta AI.


