Meta's AI Application Allows Users to Unearth Strange, Intimate Conversations of Other Users
"Navigating the Uncharted Waters: Meta AI and the Public's Private Conversations"
Curiosity sparked a query on Meta's AI platform about young women preferring older white men, igniting a discussion that left a 66-year-old single man from Iowa pondering a leap to distant Mediterranean or Eastern European lands. As intriguing as this scenario seems, it's just one of scads of apparent personal chats that can be viewed by all on Meta AI, a relatively new chatbot platform, born in April.
The "Discover" tab within the Meta AI app serves as a timeline of others' interactions, hinting at seemingly private exchanges in a swipe down the Meta AI website. Some of the queries and responses seem harmless, like travel itineraries, while others disclose locations, telephone numbers, and sensitive information, attached to user names and profile photos.
Calli Schroeder, senior counsel for the Electronic Privacy Information Center, voiced her concerns over nonchalant users sharing medical information, mental health issues, home addresses, even court case details in our interview. "All of this revelatory stuff is concerning," Schroeder noted, "as it appears people are misinterpreting the purposes of these chatbots and the privacy implications."
However, it's not clear whether Meta AI users are aware their chats are open to the public, or whether trolls have taken advantage of recent news coverage. While privacy isn't the default setting on Meta AI, users must manually opt-in to share their conversations.
A restless landlord sought guidance on evicting a tenant, while another questioned the AI about crafting an academic warning notice, detailing their school's name. Some probed inevitable corporate tax fraud in a specific location, leveraging an Instagram account tied to their identity. One individual even asked the AI to craft a court character statement, divulging personal details that implicated both the accused and themselves.
Medical concerns were rife in the queries, with users airing their struggles with bowel movements, hives, and rashes. One user confided in the AI about their neck surgery, volunteering their age and occupation. Many, but not all, accounts appeared to align with a public Instagram profile showcasing the user's identity.
Meta's spokesperson Daniel Roberts addressed our concerns in a statement, explaining that chats remain private unless explicitly shared on the "Discover" feed. However, the company didn't respond to queries regarding measures in place to protect sensitive information disclosed on the platform.
Privacy Considerations
- Exposure of Private Chats: The public availability of chats can expose sensitive information, leading to potential identity theft and harassment.
- Lack of Encryption: Unlike secure messaging apps, Meta AI chats are not encrypted, increasing the risk of unauthorized access.
- Unintended Sharing: Users may unknowingly share sensitive information if their public Instagram accounts are linked, exposing their searches to others.
Recommended Actions
- Avoid Sensitive Discussions: Users should steer clear of discussing personal or sensitive matters on Meta AI.
- Choose Wisely: Consider using more privacy-focused chatbots until Meta AI resolves its privacy issues.
[1] Knight, W. (2023). "Meta AI's New Chatbot Puts Your Sensitive Conversations on Blast". MIT Technology Review.[2] Schiller, L. (2023). "Meta AI: The Unintended Privacy Consequences of a Chatbot Gone Viral". Wired.[3] Lee, Y. (2023). "The Privacy Risks of Talking to a Chatbot: What You Should Know". Consumer Reports.
- Despite being a relatively new chatbot platform, Meta AI's lack of encryption and public availability of chats raises concerns about the exposure of private conversations, potentially leading to identity theft and harassment.
- The "Discover" tab within Meta AI's app, functioning as a timeline of others' interactions, has revealed a plethora of personal exchanges that include sensitive information like locations, telephone numbers, and medical conditions, which users may not intend to share with the public.
- Concerned parties, such as Calli Schroeder from the Electronic Privacy Information Center, suggest that users are misinterpreting the purposes of these chatbots and the privacy implications, given that Meta AI doesn't secure users' chats by default.
- Meta AI's spokesperson has acknowledged that chats are private unless explicitly shared on the "Discover" feed, but further inquiries regarding measures to protect sensitive information disclosed on the platform have not been addressed.
