Some users of Meta AI may unknowingly have their search prompts and results posted publicly on a feed accessible to everyone. The tool allows users to share their interactions, but not all appear aware of the visibility of these posts.
An internet safety expert called this a significant user experience and security issue, noting that posts can be traced back to individuals via usernames and profile pictures linked to social media accounts. This exposes potentially sensitive searches, including requests for generating scantily clad characters or solutions to academic tests.
Meta has been approached for comment. While sharing is not automatic, users who choose to post receive a warning: “Prompts you post are public and visible to everyone… Avoid sharing personal or sensitive information.” Despite this, numerous examples were found where people uploaded school or university test questions or sought help with exams.
Some publicly viewable chats feature searches for revealing images of women or anthropomorphic characters, sometimes traceable to the users’ Instagram profiles. Tech outlet TechCrunch also reported intimate medical questions posted openly, such as inquiries about treating rashes.
Meta AI is accessible via Facebook, Instagram, WhatsApp, and as a standalone product featuring a public “Discover” feed. Users can set searches to private in their account settings. The service is available in the UK via browser and in the US through an app.
In an April press release, Meta stated the “Discover” feed is “a place to share and explore how others are using AI,” emphasizing that posts only appear if users choose to share them.
Rachel Tobac, CEO of cybersecurity firm Social Proof Security, criticized the mismatch between user expectations and reality, pointing out that many do not expect their AI interactions to become public content linked to their identities. She warned this could lead to inadvertent exposure of sensitive information.