OpenAI removes private chat indexing feature

OpenAI has removed a controversial ChatGPT feature that led to some users unintentionally having their private chats appear in Google search results. The issue, identified by Fast Company, found that thousands of highly personal conversations were publicly indexed. Details included descriptions of interpersonal relationships with friends and family that could potentially be used to identify those involved. 

Users whose chats were exposed had opted in by ticking a box after choosing to share a chat. However, the opt-in text was small and lightly formatted, potentially misleading users into sharing conversations about sensitive topics like their sex lives, mental health, drug use and traumatic experiences.

While OpenAI initially defended the feature, Dane Stuckey, Chief Information Security Officer at OpenAI, confirmed the company decided it created "too many opportunities for folks to accidentally share things." While OpenAI has promised to remove the indexed content from Google, it remains unclear what role Google played. Carissa Veliz, an AI ethicist at the University of Oxford, criticised the experiment as "troubling," suggesting that tech companies use the public as "guinea pigs" to test invasive design choices.

Meanwhile, WIRED (£) reports that Security researchers have discovered a vulnerability that could allow hackers to gain access to ChatGPT user data through linked accounts via third-party platforms.

£ - This article requires a subscription.