Artificial intelligence is invited every day a little more in our daily lives, attractive by its advice and its ability to dialogue like a human. But what does our supposedly private conversations really become with tools like Chatgpt? These exchanges, supposed in particular to help and reassure, could be exposed far beyond the expected circle.
According to information revealed by the American magazine Fast Company on July 30, a simple search on Google could bring up to 4,500 conversations shared from Chatgpt. Behind this phenomenon, there is a technical detail: a user can click on “Share” to send their conversation to a close friend via WhatsApp or to record the URL for subsequent consultation. However, this action also shows it in Google search results, accessible to all.
Among the thousands of discussions thus listed, some reveal intimate stories: confidences on addiction, trauma, mental health question, revelations about his sex life or his family difficulties. If the system does not show your name on Google, the information shared during discussion with artificial intelligence is sometimes very specific and can therefore allow some to identify you.
This situation is all the more disturbing that almost one in two American states having used a chatbot to request psychological support in recent months. Many do not imagine that their secrets can one day be found in free access.
OPENAI says it has deleted the functionality
Most users were unaware that by creating this type of link, it was possible to find it in the search results, accessible to everyone. The intention was often to share the URL in a small committee, without any intention to make the thing public. Experts alert for a long time for the lack of clear information concerning the use of these services and the gap between what users believe and the real operation of the platforms.
The 1er August, Openai reacted by indicating on x having deleted “The functionality that allowed users to make their conversations visible by search engines, like Google”.
We just removed a feature from @ChatgptApp that allowed users to make their discoverable conversations by search engine, such as google. This was a short-lived experience to help people discover useful conversations. This feature required users to opt-in, first by cat picking… pic.twitter.com/mgi3lf05ua
– Danξ (@cryps1S) July 31, 2025
Thousands of conversations had already been made public and indexed. Openai also ensures in its communication on x “Work to delete the indexed content of the search engines concerned”.
This phenomenon does not only concern Chatgpt. Other platforms, such as Meta or Google Bard, encounter the same difficulties. Incidents are multiplying, proving that pedagogy on online sharing remains insufficient, even with experienced users. At a time when artificial intelligence is installed as a privileged interlocutor, thinking about its limits becomes urgent. Keep in mind that a simple click can open your privacy … to the eye of the web.
Writing note: The article has been updated with the reaction of Openai
