If your internet search history was kept online to see others, how would you feel?
This may be for some users of meta AI, without feeling, because people indicate for artificial intelligence tools – and results – are posted on a public feed.
An internet security expert said that it was “a huge user experience and safety problem” because some posts are easily ready for social media accounts.
This means that some people can inadvertently tell the world about their discoveries – such as asking AI to help AI to generate scraper characters or cheat on tests.
Meta says that chat is private by default, and if users make a post public then they can choose to withdraw it later.
Before sharing a post, a message pops up that says: “You are posted signs public and appear for everyone … avoid sharing personal or sensitive information.”
However – Given the personal nature of some questions – it is not clear whether users understand whether their discoveries are being posted in a public “Discover” feed on the meta AI app and website, and they can be detected by their other social accounts through user names and profile paintings.
The BBC found several examples of many examples of people uploading pictures of school or university test questions, and asked Meta AI for answer.
One of the chats is the title “Generative AI easily deal with mathematics problems”.
Another user interaction was posted publicly about searching questions around their gender and should they infection.
There were also discoveries for women wearing very few clothes and anthropomorphic animal characters.
A discovery, which can be detected back to a person’s Instagram account due to their user name and profile drawing, asked the meta AI to generate the image of an animated character wearing only underwear.
Meta AI, launched earlier this year, can be accessed through its social media platform Facebook, Instagram and WhatsApp,
It is also available as a standalone product that has a public “discover” feed.
Users can choose to make their discoveries private in their account settings.
Meta AI is currently available through a browser in the UK, while in the US it can be used through an app.
One in Press release from April Which announced the meta AI, the company said that “a place would be a place to feed, share and find out how others are using AI”.
“You are under control: until you choose to post it, your feed is not shared anything,” said this.
But Rachel Tobac, US Cyber ​​Security Company Social Proof Security Chief Executive Officer, Posted on X Saying this: “If a user’s expectations do not match the reality, you have found yourself a huge user experience and safety problem.”
He said that people do not expect to make their AI chatbot interaction public on social media feeds.
“Because of this, users are inadvertently posting sensitive information for public feed associated with their identity,” he said.
