OpenAI has removed the feature that allowed to show joint conversations in ChatGPT in search results. The “short-term experiment” was based on the option to create links in the chatbot. Following complaints, OpenAI’s Chief Information Security Officer Dane Stuckey said the company was working to remove the chats from search engines.
Public outrage was sparked by a Fast Company article published earlier this week (via Ars Technica). Fast Company claimed to have found thousands of ChatGPT conversations in Google search results. The indexed chats did not contain any explicit identifying information. But in some cases, their content was reported to contain specific details that could point to the source.
To be clear, this was not a hack or leak. It had to do with a checkbox that users could check when creating a public URL leading to a chat. In the pop-up window for creating a public link, the option “Make this chat searchable” appeared. A more direct explanation (“allows it to be shown in search engines”) appeared in smaller, grayer font below. Users had to check this box for the chat to be indexed.
You may wonder why people who create a public link to a chat have a problem with its content being public. But Fast Company points out that people may have created URLs to share in messaging apps or as an easy way to return to the chat later. Regardless, the possibility of public discovery is now gone.
In a Fast Company report, Stuckey defended the feature’s labeling as “clear enough.” But as the outrage grew, OpenAI relented. “Ultimately, we believe that this feature created too many opportunities for people to accidentally share something they didn’t intend, so we are removing this option,” Stuckey announced on Thursday.