Meta AI Public Feed Exposes User Searches
It's a startling thought: what if your private internet searches suddenly became public? For some users of Meta AI, this scenario might be an unintended reality. Prompts given to the artificial intelligence tool, along with the AI's responses, are reportedly appearing on a public feed, potentially without the users fully realizing the implications.
Are Your Meta AI Searches Unintentionally Public
An internet safety expert has described this situation as "a huge user experience and security problem." The main concern is that some of these public posts can be easily traced back to users' social media accounts. This means individuals could be unknowingly sharing sensitive queries with the world, such as asking Meta AI to create images of scantily clad characters or for assistance with cheating on exams.
Meta maintains that its AI chats are private by default. According to the company, if users decide to make a post public, they are presented with a pop up message. This warning states, "Prompts you post are public and visible to everyone... Avoid sharing personal or sensitive information." Meta also notes that users can retract these public posts later.
The Disconnect User Awareness vs Public Feeds
Despite Meta's assurances and warning systems, the private nature of many queries appearing on the public "Discover" feed raises questions. It's unclear if users truly understand that their interactions are being shared publicly on the Meta AI app and website. More alarmingly, these public posts can sometimes be linked to their other social media profiles through shared usernames and profile pictures.
The BBC uncovered several instances demonstrating this issue. These included users uploading photos of school or university test questions and asking Meta AI for solutions. One such public chat was titled "Generative AI tackles math problems with ease." In another publicly posted conversation, a user explored personal questions about their gender identity and potential transition. There were also numerous searches for images of women and anthropomorphic animal characters in minimal clothing. A striking example involved a user, identifiable via their Instagram username and profile picture, asking Meta AI to generate an image of an animated character in underwear.
Metas AI Platform and the Discover Feature
Meta AI, which launched earlier this year, can be accessed through popular social media platforms like Facebook, Instagram, and WhatsApp. It's also available as a standalone product featuring the public "Discover" feed. Users do have the option to set their searches to private within their account settings. Currently, Meta AI is accessible in the UK via a browser and in the US through a dedicated app.
In an official announcement from April detailing the Meta AI launch, the company mentioned the "Discover feed" as "a place to share and explore how others are using AI." The press release further stated, "You're in control: nothing is shared to your feed unless you choose to post it."
Expert Concerns A Mismatch in Expectations
However, these explanations haven't satisfied all critics. Rachel Tobac, chief executive of the US cybersecurity company Social Proof Security, voiced strong concerns. In a post on X, she stated, "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."
Tobac elaborated that individuals typically do not expect their interactions with an AI chatbot to be made public on a feed commonly associated with social media content sharing. "Because of this, users are inadvertently posting sensitive info to a public feed with their identity linked," she added. This highlights a critical gap between how users perceive AI tools and how they might actually operate regarding data visibility.
