I never thought I’d download Meta AI on my iPhone. After all, people have been mad for over a year that you can’t turn off Meta AI on Facebook and Instagram: Why would you want a dedicated app for this?
Then, I saw the headlines from TechCrunch, Wired, and Business Insider, among many others, that sharply criticize the app’s approach to privacy and security. That’s because Meta AI isn’t Meta’s take on ChatGPT, Gemini, or Claude. Instead, it’s half-chatbot, half-social media platform, where your requests and questions can be shared with the rest of the Meta AI community.
To be clear, your Meta AI interactions aren’t shared by default. You do need to choose to post your queries to the social aspect of the app. Should you choose to do so, your requests are posted to a public Discover feed for all with the app to see. That invites users who want to share their AI creations, of course, as well as trolls who want to spam the feed with silly or offensive requests and generations. But what’s more concerning about the feed, however, is that it hosts posts from users who clearly did not understand they were posting their chats publicly.
My first impression when scrolling through this feed was that it does seem like most users are in on the social aspect of the app. Some posts seem geared towards a public audience, with users commenting as they would on an Instagram of Facebook post. And, in fact, most of these posts are pretty harmless: a Maltese dog swimming in a pool; killer clowns from outer space; Nartuo and Deku clashing in an epic multiverse battle; and lots of anthropomorphic animals…so many anthropomorphic animals.
Even some of the personal posts are fine with an audience: I saw a screenshot of someone’s sleep stats presumably tracked from an Apple Watch, and the user was asking Meta AI to analyze the stats and report what it thought. The user was then responding to the comments, including to a user who said “audit looks good” and one who inquired into whether the user sleeps with “these” on—”these” presumably meaning a smart watch? Why you’d want to start a public discussion about your sleep habits with strangers is beyond me, but to each their own.
But every once in a while, you come across something that was clearly intended to remain private—at least between the user and Meta—and is now visible to me. I saw someone post a very close-up selfie asking for a beautiful yet realistic makeover. They didn’t seem happy with the digital, carnival-like makeup the AI used, because they asked once again for it to look realistic. (It didn’t.) Someone asked for a “skinny cute girl anthro lion in pink striped socks, poledancing in the club.” (I can report Meta AI did animate this.) The same man then posted an image of himself with two women kissing in the background (those women were added by Meta AI in a separate request) and asked Meta to keep adding more and more women to the background. The sole comment I saw read: “these are publicly posted, my guy.”
It’s not just that these posts are public: They’re tied to your Meta account, which anyone can tap through to see your entire posting history. One user who asked Meta to generate “muddy bikinis and passionate kisses” also asked Meta AI what to do about a number of red bumps on their inner thigh. Sorry you’re going through that, but, also, why do I and the rest of the Meta AI app know? I’m going to go out on a limb and assume you didn’t mean to share that—unless you’re just messing with us.
In general, Meta has a poor reputation when it comes to privacy. (This is the company that allowed 87 million accounts to be exposed to Cambridge Analytica, after all.) But this app is a privacy and security nightmare.
Taking a look at the iOS App Store’s app privacy report card, Meta AI scrapes a ton of your data, including health and fitness, financial information, contact information, browsing history, usage data, your location, contacts, search history, sensitive information, and “surroundings,” which I didn’t even know was a metric. (According to Apple, it means “environment scanning,” including “mesh, planes, scene classification, and/or image detection of the user’s surroundings.”)
But digging through the settings, there are some terrible defaults here. First of all, Meta will suggest your publicly-shared prompts on other platforms, like Facebook and Instagram. (You can turn this off from Meta AI’s in-app settings by tapping your profile icon, then from Data & privacy > Suggesting your prompts on other apps.) Meta AI also keeps “background conversations” enabled, which basically listens to you when you leave the app or put your phone to sleep in case you want to keep chatting with Meta AI at any time. No thanks. (You can turn this off from Settings > Data & privacy > Voice.)
There are also some serious security implications here. One person shared a photo of their computer (one taken with a camera, not a screenshot, mind you), which displayed a warning that their Facebook account would be disabled in 169 days. As it turns out, Facebook had banned their account, but the user had appealed and was confused about what to do next. The user shared their full name, and asked if the bot could “talk to that AI about my appeal?” They then, without prompting, shared what kind of business they run, which was enough information for me to find both their LinkedIn and Instagram accounts. I’m pretty confident this person had no idea their requests to Meta AI were posted publicly, and my guess is if they knew, they wouldn’t be sharing these details so lightly—or conversing with Meta AI at all.
Too many people seem unaware that the posts they share with Meta AI of themselves with friends and family, or the deeply sensitive questions they entrust to the bot, are now public to the community of users scrolling through the Discover tab. Of course, there’s always the chance that any one of these posts is an elaborate troll, especially with the recent media attention on Meta AI’s app. But something tells me the person asking for help with their account ban is a real person, who had no idea their desperate conversation with a Meta product would end up on a public feed, let alone in this article.
Make all your public Meta AI posts private
If you have been posting to the Meta AI app without realizing it, or just regret all of your public posts, you can make them all private in one fell swoop. To do so, head to Settings > Data & privacy > Manage your information. Tap “Make all public prompts visible to only you,” then “Apply to all” to privatize all posts. Or, you can tap “Delete all prompts,” then “Delete all” to get rid of them for good.
