OpenAI quickly rolled back a new feature that allowed users to make private conversations with ChatGPT "discoverable" after the launch was marred by concerns of accidental oversharing.
"We just removed a feature from @ChatGPTapp that allowed users to make their conversations discoverable by search engines, such as Google," Dane Stuckey, OpenAI's chief information security officer, announced in a social media post on Thursday.
He described the opt-in feature as "a short-lived experiment to help people discover useful conversations" but added that the company removed it in part because of OpenAI's focus on "security and privacy."
"Ultimately we think this feature introduced too many opportunities for folks to accidentally share things they didn't intend to, so we're removing the option," Stuckey's post said. "We're also working to remove indexed content from the relevant search engines. This change is rolling out to all users through tomorrow morning."
On Wednesday, Fast Company was the first to report that some ChatGPT conversations were being indexed by Google. The rollback came shortly after the newsletter writer Luiza Jarovsky posted on X on Thursday that sensitive conversations with ChatGPT were being made public. Jarovsky wrote that when using the chatbot's sharing feature, users were opening up their exchanges to being indexed by Google.
The feature required users to take steps to share their conversations — including ticking a box to "make this chat discoverable" — and indicated they would "be shown in web searches." Chats made public were anonymized, reducing the risk that someone may be personally identified by their conversations with the bot.
Still, users responding to Jarovsky's posts highlighted the concern that some people may absent-mindedly check the box to share without reading the fine print, opening themselves up to sharing embarrassing or private information.
In her initial X post, Jarovsky found examples of users discussing their fears with ChatGPT. Her comments were filled with more personal examples, including discussions of harassment and makeshift therapy sessions.
Jarovsky and representatives for OpenAI didn't immediately respond to requests for comment.
