Some Chatbots Revealed Personal Information More Easily Than Others

CNET found uneven privacy behavior across AI chatbots: Grok quickly surfaced phone numbers and home addresses, while ChatGPT sometimes returned partial personal details.

Gemini was more restrictive and refused to provide private information, showing major differences in how chatbots handle sensitive requests.

A Cornell study cited in the piece says several major AI companies use user inputs to train models by default unless users opt out.

Some companies may retain prompt data indefinitely, underscoring broader concerns about how personal information is stored, reused, and exposed.