Thousands of private ChatGPT discussions have surfaced in Google search results because of the chatbot’s “Share” option, which has now been withdrawn following criticism.
Fast Company highlighted that ChatGPT users might have inadvertently made their interactions public and searchable. The report identified almost 4,500 ChatGPT discussions in Google search results, some touching on personal and sensitive themes such as mental health and relationships. Thankfully, these publicly accessible conversations did not disclose the identities of the users.
How did these interactions wind up online?
In the past, ChatGPT users had the ability to share chats by making them public, akin to Google Doc sharing configurations. Users would get a public link to distribute to others. An option existed allowing users to set the post as “discoverable,” specifically by Google, whether they understood it or not.
When generating a shareable link, a pop-up notified users: “A public link to your chat has been created.” A checkbox titled “Make this chat discoverable” was displayed, accompanied by a warning: “Allows it to be shown in web searches.”
By checking this box, users permitted their conversations to be indexed by Google, resulting in their appearance in search results.
Following Fast Company’s report, OpenAI eliminated the feature, with a company executive describing it as a “short-lived experiment.”
OpenAI Chief Information Security Officer Dane Stuckey detailed on X how the feature functioned and where it failed.
Even though users had to opt in for their chats to be public, the company concluded that the risk of user error was too significant.
As Mashable noted, OpenAI is obligated to retain user conversations, even those deleted, due to an ongoing lawsuit initiated by the New York Times. OpenAI must keep all conversations indefinitely, though this requirement does not extend to ChatGPT Enterprise or ChatGPT Edu customers.
While ChatGPT users may activate a “Temporary Chat” feature reminiscent of incognito mode, chat data could still be preserved.
Disclosure: Ziff Davis, Mashable’s parent company, lodged a lawsuit against OpenAI in April, alleging copyright violations in the training and operation of its AI systems.