• by Laurie Sullivan  @lauriesullivan, Yesterday

Security and financial experts have noted that Meta AI makes private conversations public when users try to post messages to its social feed without understanding exactly how the platform works. The conversations are searchable, which could turn into a privacy nightmare for the company and its users.

The supposed glitch and apparent ignorance on the part of the user present a major security problem for users. It means people are unwittingly airing personal information online.

This is the type of information they would not want others to know or see because they are posting sensitive data to a public feed with their identity linked, Rachel Tobac, CEO at online security firm SocialProof Security, wrote in a post on X.

These posts might include questions the AI is asked about such personal matters as “feeling shame with disabilities,” “exact medical issues,” or “private upcoming travel plans.”

Justine Moore, partner at venture capital firm Andreessen Horowitz (a16z), has also seen that “wild things are happening on Meta’s AI app.”

She wrote that the “feed is almost entirely boomers who seem to have no idea their conversations with the chatbot are posted publicly," and posted an anonymized picture of a chat she saw.

Moore explained that these posts are not being shared by default. People need to hit a “share” button, but many users seem to think they are posting to a private journal.

The misunderstandings about how to use Meta AI are resulting in communications being made public, such as a man who tried to write a poem to his girlfriend.

“You can hear a 6-min audio clip,” Moore wrote, posting a written transcription of the piece.

One man tried to use Meta AI to find a woman with large breasts and when the message did not post on his behalf in local Facebook groups, he asked the bot to delete his phone number. Instead, it was shared publicly.

It’s not clear exactly how this occurs without the user’s knowledge or permission. The BBC explains that if people choose to share a post, a message pops up that says: "Prompts you post are public and visible to everyone... Avoid sharing personal or sensitive information."

The BBC reported finding several examples of people uploading photos of school or university test questions, and asking Meta AI for answers. And TechCrunch found examples of asking the Meta AI app for ways to evade taxes and how to write a character reference letter for an employee facing legal troubles.

The Meta AI app has only been downloaded 6.5 million times since it debuted on April 29, according to Appfigures, an app intelligence firm.

Trust in Meta AI could become a major issue for the company, despite is recent investment of nearly $15 billion in Scale AI.

Alexandr Wang, Scales’ founder, announced the investment on Thursday and said he will join Meta to work on its AI efforts, but also serve as a director on the Scale Board of Directors.

The board has appointed Jason Droege, chief strategy officer, to serve as the company’s interim CEO. Droege joined Scale in September 2024 with more than 20 years of experience building and leading technology companies. Before joining the company he co-founded Uber Eats and was a venture partner at Benchmark, an investment firm focused on early-stage venture investing in consumer, marketplaces, open-source, AI, infrastructure, and enterprise software.

MediaPost reached out to Meta for comment, and will update the article with any additional information.