OpenAI said it is working to remove users ChatGPT chats from public search engine listings, following a report that highlighted that many chats had become discoverable and contained sensitive personal information.
The chats were indexed on Google and other search engines after users clicked a “Share” button on ChatGPT to create a link to provide access to the chat to other people, Fast Company reported.
It was unclear whether people were aware that they were sharing their conversations with the chatbot to the broader internet when clicking the “Share” button, which has now been removed.
Sensitive chats
In some cases users may have been creating the link to refer back to it later themselves, or to send it to a relatively small group of people.
Conversations found via public search engines included personal details, including people’s experiences of addiction, physical abuse, mental health issues, and fears that AI models were spying on them, the report said.
While the chats did not explicitly identify users, in some cases specific details could allow the user’s identity to be discovered, the magazine reported.
OpenAI chief security officer Dane Stuckey called the feature a “short-lived experiment” and said the start-up was working to remove the chats from search engines.
He defended the feature’s labelling as “sufficiently clear”, but acknowledged people might not have been aware of the implications of using it.
“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,” he said in a statement provided to Silicon UK.
Usage of ChatGPT has been growing rapidly, with the company posting recurring annual revenue of $13 billion (£9.8bn) as of the end of July, up from $10bn in June, with the goal of exceeding $20bn by the end of the year, according to The New York Times‘ Dealbook.
The newspaper reported that the start-up has raised $8.3bn in a new funding round that values it at $300bn, part of a $40bn funding round announced in March.
The round, led by SoftBank with a $30bn investment, included another $2.5bn from venture capital firms.
Intense competition
The latest tranche was led by Dragoneer Investment Group with some $2.8bn, which the newspaper said could be one of the largest-ever single investments from one firm.
The round was five times oversubscribed, the paper reported.
The start-up faces competition from the likes of Amazon-backed Anthropic, which said software programmers now prefer its models over those of others.
It said it had found that developers from OpenAI itself were using Anthropic’s coding tools as they prepare the GPT-5 model, which Anthropic called a violation of its terms of service, as a result of which it revoked OpenAI’s access to its developer tools.
OpenAI said it was standard practice to use competing AI tools to benchmark progress and improve safety, in remarks earlier reported by Wired.
Anthropic earlier revoked API access to start-up Windsurf at a time when it appeared that Windsurf might be bought by OpenAI, in a deal that later fell through.