A researcher compiled nearly 100,000 publicly shared ChatGPT conversations indexed by Google, highlighting significant privacy issues. The content ranges from sensitive materials, such as non-disclosure agreements and confidential discussions, to mundane queries like relationship advice and LinkedIn post requests. OpenAI acknowledged the dataset, confirming the removal of a feature that allowed user-generated content to be discoverable by search engines. This feature was briefly tested for sharing useful conversations, requiring user opt-in for visibility.
The dataset includes everything from sensitive documents like non-disclosure agreements to users discussing personal relationship issues and professional inquiries, illustrating significant privacy concerns.
OpenAI did not dispute the figure of around 100,000 indexed chats. A feature to make conversations discoverable by search engines was removed recently, after being an experiment.
Collection
[
|
...
]