Private ChatGPT discussions, unintentionally exposed online by Google
OpenAI has announced the removal of a feature in its popular chatbot, ChatGPT, that allowed users to make their conversations publicly searchable on search engines like Google. The feature, which was described as a short-lived experiment, has been discontinued due to concerns about accidental privacy breaches.
Dane Stucsky, OpenAI's Chief Information Security Officer, made the announcement. He stated that the feature provided too many opportunities for users to accidentally share things they didn't want to, which conflicted with OpenAI's emphasis on security and privacy.
The feature, which required users to actively consent by checking a box, allowed private conversations with ChatGPT to be made discoverable on search engines. This led to users inadvertently making private or sensitive conversations publicly searchable, exposing personal information such as names, locations, emails, and work details.
Examples of sensitive conversations that were made public include users discussing their fears, stressors, and even improvised therapy sessions. Jarovsky, a technology analyst, highlighted these concerns, stating that this could potentially lead to sensitive content being accessed by anyone via search engines.
OpenAI is actively working to remove already indexed content from search engines to mitigate the exposure. The change will be rolled out to all users by tomorrow morning.
Users have expressed concern that some people might carelessly check the box without reading the fine print and accidentally share embarrassing or private information. This move by OpenAI is aimed at ensuring the privacy and security of its users, demonstrating the company's commitment to these values.
The feature, intended to help people discover useful conversations, is no longer available on the ChatGPT app. OpenAI encourages users to be vigilant about their privacy settings and to only share information that they are comfortable with being made public.
[1] TomsGuide Article [2] SimonWillison Article [3] TechCrunch Article [4] Business Insider Article
- Dane Stucsky, OpenAI's Chief Information Security Officer, emphasized that the discontinued feature, which allowed private conversations with ChatGPT to be made discoverable on search engines, presented too many opportunities for users to inadvertently share sensitive information of their own, such as names, locations, and work details, thus conflicting with OpenAI's focus on security and privacy.
- Jarovsky, a technology analyst, said that the feature could potentially lead to sensitive content being accessed by anyone via search engines, highlighting concerns about the unintentional exposure of personal information of discussions that were meant to be private, including fears, stressors, and even improvised therapy sessions.
