A recent incident involved OpenAI removing a sharing feature after users unintentionally exposed private ChatGPT conversations. The 'share' button's misleading design allowed users to create public links, inadvertently making their discussions accessible online. OpenAI's efforts to de-index these conversations from search engines were partially successful, yet over 110,000 exchanges remain available through Archive.org. One disturbing example involved a lawyer for an energy corporation discussing strategies to displace an indigenous tribe to develop a hydroelectric project, revealing the exploitative attitudes of some users.
Users unknowingly exposed their private ChatGPT exchanges by using a 'share' function that made conversations public and indexable by search engines.
Over 110,000 conversations remain accessible via Archive.org, containing alarming content including unethical discussions on displacing indigenous communities for corporate profit.
An Italian-speaking lawyer discussed how to negotiate the lowest price for displacing an Amazonian community for a hydroelectric project, expressing exploitative attitudes.
The leak resulted from poor design and user errors, not malicious hacking, leading OpenAI to remove the 'discoverable' option and de-index conversations.
Collection
[
|
...
]