Meta AI is just another AI model among many popping up from existing and popular companies. What’s drawing people in, though, is the AI chatbot’s direct tie into Meta’s social apps people are already using to post, like Facebook and Instagram. It turns out that blurring that line between social and chatbot means people are asking a lot of personal questions, and they are all out in the open.
There’s a slew of companies that have added AI to existing apps, and Meta‘s Llama model has done some significant improvements over the past year or so, though it isn’t quite up to par with OpenAI’s o3 or Google’s Gemini Deep Research models. What it is good at is entertainment, and the Meta AI app is providing some for everyone.
The basis of the Meta AI app is simple. There is a text box that lets you directly communicate with the model, and there are a couple of options as to how that happens. One, text works fine. Two, users can chat vocally with the model. Doing so records your voice and interporates it as speech, like any other live AI model.
The kicker is this: these conversations can be made public relatively easily. After you’ve had the conversation, you can hit the share button in the top corner. In doing so, you can agree to post your conversation to Meta AI’s Discover feed.
It seems like a lot of users are posting their conversations without realizing it.
Scrolling through the “Discover” page reveals that a lot of users are seemingly hitting that post button without a second thought, whether by mistake or on purpose. While using the app for my limited time, I’ve seen some absolutely wild inquiries.
Those posts range from older users asking Meta AI very personal medical questions to people legitimately using AI to write character testimonies for their friends who are standing trial in an active court case. A lot of them also seem to be incoherent nonsense and it’s truly shocking that Meta’s Llama model knows what to do with that information.
One specific interaction contained an hour-long audio clip of a user who seemed to have Meta AI on retainer for a long drive. The posts consisted of questions the user seemingly didn’t actually want the answer to and some rather inappropriate questions that likely were the result of some previous illegal activity – the kind of thoughts that might enter your head but shouldn’t leave, especially not on something directly tied to your Facebook account. All of this stemmed from this user’s actual voice recording, which feels all too personal.
Becuase that’s just it, Meta AI has a direct tie to Facebook and Instagram. These are not anonymous searches with Meta’s AI model, and they’re being broadcast as posts for others to leave comments or reactions on. To a lot of users’ credit, many comments are simple ones that remind the original poster that their conversation is public and they should consider taking the post down.
Most of the content in the above images has been redacted because these queries are just too personal to be public. It doesn’t take long to find something off-putting on the discover page, or perhaps that needs to be brought up to a licensed therapist.
One could argue that it’s just how these social sites work, and that might be true. In any case, Meta AI’s public feed needs to be harder to post to. If not, these personal conversations will keep being made public.
Follow Andrew: Bluesky, Twitter/X, and Threads
H/T: Business Insider
More on AI:
- Report: Google and OpenAI reach cloud deal despite ChatGPT vs Gemini rivalry
- Gemini coming to Google Docs for Android
- It’s now just the ‘Gemini’ app as ‘Gemini Pro’ brand goes away
FTC: We use income earning auto affiliate links. More.
Comments