More than 370 thousand conversations on the Grok AI platform were published on the Grok website and indexed by search engines, 9to5mac reported.
The xAI company not only publicly released conversations but also images, tables, and other documents uploaded by users.
Links created through the "share" button on Grok were made accessible to search engines. This means that not only the people the link was sent to but the entire internet users can see them.
According to Forbes, users were not warned that conversations and uploaded documents would be publicly accessible.
By clicking the “share” button on Grok, any conversation is posted on the site without warning.
When searching for Grok conversations on Google, it was found that over 370 thousand conversations between users and Grok appeared. These include topics ranging from simple tweets to descriptions of terrorist attacks and hacking crypto wallets.
Among the conversations analyzed by Forbes, personal questions related to medicine and psychology, names, passwords, and other confidential information were also disclosed. Uploaded files and documents were also made public.
Some conversations were found to violate the policy set by Grok itself. It provided users with detailed instructions on illegal drug manufacturing, creating harmful software, suicide, and even plans to kill Elon Musk.
This is not the first time such an incident has occurred. Previously, some ChatGPT conversations appeared on Google, but in that case, the user had consented.
However, this situation is particularly ironic because Elon Musk made unfounded confidentiality claims last year about the Apple and OpenAI partnership.
Read 'Zamin' on Telegram!
