Users of Meta AI are unknowingly broadcasting their most personal queries to the world, creating what one cybersecurity expert calls “a huge user experience and security problem.”
The artificial intelligence tool, which launched earlier this year across Facebook, Instagram, and WhatsApp, features a public “Discover” feed where conversations can appear without users fully understanding the implications. Some of these posts are easily traceable back to people’s social media accounts through matching usernames and profile pictures.
What’s Being Shared with Meta?
The BBC investigation uncovered several concerning examples of what people are inadvertently making public. Students have uploaded photos of test questions from school and university, asking Meta AI to provide answers. One such conversation was even titled “Generative AI tackles math problems with ease.”
More sensitive searches have also appeared on the public feed. One user’s deeply personal conversation explored questions about gender identity and whether they should transition. Others have requested images of scantily-clad characters, including searches for women and anthropomorphic animal characters in revealing clothing.
Perhaps most troubling, some of these requests can be directly linked to users’ real identities. One search that asked Meta AI to generate an image of an animated character in underwear was traceable back to the person’s Instagram account through their identical username and profile picture.
How does the System of Meta Work?
Meta AI operates across the company’s social media platforms and is also available as a standalone product with its own public feed. Users can access it through browsers in the UK, while Americans can download a dedicated app.
Before sharing a post publicly, the system does display a warning message stating: “Your post is public and visible to everyone… Avoid sharing personal or sensitive information.” However, given the private nature of many queries found on the feed, it appears users either don’t understand this warning or don’t realize their searches might be shared.
The company has said that chats remain private by default, and users can withdraw posts after making them public. There’s also an option in account settings to keep searches private entirely.
The Problem According to Experts
Rachel Tobac, chief executive of cybersecurity firm Social Proof Security, highlighted the core issue in a post on X: “If a user’s expectations about how a tool functions don’t match reality, you’ve got yourself a huge user experience and security problem.”
She explained that people naturally expect their conversations with AI chatbots to remain private. When these interactions appear on what looks like a typical social media feed, users can end up “inadvertently posting sensitive info to a public feed with their identity linked.”
This disconnect between user expectations and reality creates significant privacy risks, especially when personal information can be connected to real identities through social media profiles.
The Position of Meta
When Meta revealed the AI tool in April, the company highlighted user control when it issued its press release. “You’re in control: nothing is posted to your feed unless you opt to post it,” they said, characterizing the Discover feed as “a space to share and see how other people are applying AI.”
The company asserts that users have choices of making their interactions private and deleting posts after posting them publicly.
This is an example of a larger problem in the age of AI: assisting users in understanding how their data and activity may be used or shared. As more AI utilities are built into social media, there can be ambiguity between private search queries and public content.
The instances discovered by the BBC indicate that users mostly view Meta AI as a personal assistant or private search engine without knowing that their questions may end up on an open feed accessible to everybody. This ignorance has drastic implications for people’s privacy, academic honesty, and individual security.
For users of Meta AI, the key takeaway is clear: check your privacy settings and think carefully before hitting share. What feels like a private conversation with an AI assistant might actually be broadcast to the world.
