Meta AI Adds an Extra Step to Prevent Embarrassing Overshares

That delicate conversation about your wife, or a call for help about your daughter — or the weird pocket dial that recorded a conversation with your friend — are all less likely to accidentally end up on the public feed of Meta’s AI app.

Mark Zuckerberg’s company is now giving users a new pop-up warning before they share something to Meta AI’s “Discover” feed.

This comes after Business Insider reported last week that the feed was full of personal information that appeared to have been accidentally shared — things like a post that showed a parent asking the AI for advice about a wayward daughter, the audio of a pocket dial where someone chatted to coworkers about their working hours, and the text of an employee seeking legal advice about their employer.

Meta didn’t respond to my request for comment Monday when I asked about the new pop-up.

After BI’s story last week, the BBC, The Washington Post, Wired, TechCrunch, and other media outlets ran their own stories on the apparently accidental oversharing.

The new warning message, which was visible Monday, shows up after you hit “share” on a chat in the Meta AI app. The message says, “Prompts you post are public and visible to everyone. Your prompts may be suggested by Meta on other Meta apps. Avoid sharing personal or sensitive information.”

After that warning, the “Post to feed” button is disabled until you tap once more on the middle of the screen. Only then can you post to the public Discover feed.

Meta AI’s feed had eerily personal information

Meta AI, the stand-alone AI chatbot, has its “Discover” feed for people to share images they created with AI or to post their chats with the bot. These chats aren’t public by default; you’ve always had to click “share” and then “post” to send your chats to the Discover feed.

Clearly, it seems some users didn’t understand what they were doing under the previous system. (One man I talked to last week said he hadn’t meant to post his conversation with the chatbot on car repair.)

When the app first launched in late April, I wrote about the problem of accidental public posts on Meta AI. Checking back last week, I noticed that the problem seemed even worse — there were now way more posts that seemed to be clearly shared by mistake. Some of these included intimate personal issues, like asking for help drafting a letter to a judge about child custody, medical issues, and relationship quandaries.

People on X also found examples of people posting tax information, or seeming not to even understand what they were using, like a person who asked what would happen if they put a deep heat cream on their private parts, but then asked the bot to keep the conversation private (on a public feed). Hmm.

Meta AI’s new pop-up warns people before posting

Now, of course, it’s entirely possible that some of these embarrassing purient posts were secretly intentional. (The person who asked Meta AI to make him a variety of images of celebrities doing strange things, including Ana de Armas tucking Spider-Man into bed, may have wanted people to see his stuff.)

Another change in the Discover feed on Meta AI I noticed on Monday: It’s now almost exclusively images. Last week, and the times I’d checked in on the app since it began in late April, the feed had a hefty peppering of text- and audio-based posts, like those seemingly accidental pocket dials.

Meta also didn’t respond to BI if the apparent tweak to the Discover feed was intentional — like if things other than images have been buried or disappeared — or if the change is a potential natural consequence of the new extra tap needed to share something to the public feed. Or if it’s something else altogether.

I’ll update this story if Meta gets back to me, but for now, I’m glad I don’t have to close my eyes, or cringe, or flinch (as much) when I scroll through the Meta AI Discover feed.