Your Meta AI prompts might be public – here's how to check

If you use the Meta AI app, be careful you’re not accidentally sharing anything personal. Think that’s something you’d never do? Tell that to the many, many people who are apparently exposing their private prompts on the app.

Recently, users have been reporting that the app’s Discover feed has been filled with people sharing posts that are clearly not intended for the public: sensitive medical issues, admissions of crimes, relationship advice with revealing details, and more.

When I downloaded the app to see for myself, posts seemed fairly normal at first. A bit weird, but normal. I saw AI-generated images and videos of fantasy scenes, landscapes, animals, and the occasional political figure. 

Also: Forget AGI – Meta is going after ‘superintelligence’ now

But then, about 20 seconds in, I saw a photo request for somewhat adult pictures under an account with someone’s real name. I noticed the photo had quite a few comments, so I tapped to see more. The comments were follow-up queries from the person asking for progressively more adult scenarios, and eventually comments from hundreds of people mocking this person who almost certainly didn’t know this was public.

Meta did not immediately respond to a request for comment.

As I scrolled, more than 95% of the posts I saw were pictures or videos people wanted to share. However, I would see an occasional post that probably wasn’t: someone asking for job interview help with details about the specific job, someone asking for details about what constituted extortion, and someone drafting an apology letter to their husband. I also saw people asking Meta AI to restyle photos that didn’t seem to be intended for the public, too. 

Other people have seen much worse. I didn’t see anything wildly embarrassing, but I probably would have had I kept scrolling. 

Fortunately, there’s not a search button on the app, so you can’t go specifically looking for people’s accidental shares.

So what’s going on?

How people are accidentally sharing prompts to Meta AI

Asking Meta AI a prompt doesn’t make it public. But there are two problems specific to the Meta AI app that make this an issue.

One is that when you create your Meta AI account, it defaults to your Instagram name, so many people have their actual names and pictures on their profiles. 

Also: Meta delays ‘Behemoth’ AI model, handing OpenAI and Google even more of a head start

A second is that everything is public by default on the Meta AI app. It does have a social media component, so the ability to share and be seen should be expected, but it seems like something you should have to opt-in to.

Sharing something publicly is a two-step process involving tapping “share” and then “post.” Still, people are clearly accidentally posting prompts they almost certainly don’t want public. The most likely explanation seems to be that people are intentionally clicking “post,” but think it’s some sort of personal journal.

How to make sure you’re not sharing personal prompts on Meta AI

If you’re using the Meta AI app and don’t plan on sharing anything publicly, here’s how you can make sure you don’t accidentally do so.

• Tap your profile picture on the top right corner of the app
• Tap “Data & privacy”
• Tap “Manage your information”
• Tap “Make all your prompts visible to only you”

From there, you’ll see a notification asking if you want to make all your prompts, including already posted ones, only visible to you. Choose “Apply to all.” There’s also an option to delete any posts you’ve already made.

Get the morning’s top stories in your inbox each day with our Tech Today newsletter.