How do you feel if your internet search history is placed online for others to see?
It could be happening to some users of Meta AI without them realizing it, as people's prompts (and results) for artificial intelligence tools are posted in public feeds.
One Internet safety expert said that some posts are “a huge user experience and security issue” as they can be easily traced to social media accounts through usernames and profile photos.
This means that some people may not unconsciously want the world to know what others are looking for.
Meta has been contacted for comment.
The process is not automatic, but it is not clear whether users know that searches are posted to public feeds of meta AI apps and websites.
If people choose to share their posts, they will receive the message “The post will be published and prompted to make it visible to everyone… Do not share personal or sensitive information.”
The BBC has uploaded photos of school and university test questions and found some examples of people seeking answers from Meta AI.
One of the chats is titled “Generating AI Easy to Tackle Mathematical Problems.”
There was also a search for characters from women with little clothing and personified animals.
One search that allows you to return to a person's Instagram account for username and profile photos asked Meta AI to generate an image of animated characters lying outside wearing only underwear.
Meanwhile, Tech News Outlet TechCrunch reports examples of people posting intimate medical questions, including how to deal with rashes on the inside of the thighs.
Launched earlier this year, Meta AI is accessible through social media platforms on Facebook, Instagram and WhatsApp.
It can also be used as a standalone product with a published “discovery” feed.
Users can choose to make their search private in their account settings.
Meta AI is currently available in the UK via browser, but in the US via the app.
In an April press release that announced Meta AI, the company said there is a “discovery feed, a place to share and explore how others use AI.”
“You're in control: there's nothing to share with your feed unless you choose to post it,” it said.
However, Rachel Tback, CEO of Social Proof Security for a US cybersecurity company, said, “The user's expectations about tool functionality not being consistent with reality have led to user experience and security issues.”
She added that people don't expect AI chatbot interactions to be published in feeds that are normally associated with social media.
“This is why users post careless information on public feeds with their identities linked,” she said.
