Study Finds Chatbots Like ChatGPT Can Inadvertently Infer Users' Private Details

Chatbots like ChatGPT can infer sensitive personal info about users from subtle patterns in conversations.

This is likely an unintentional consequence of how chatbot AI models are trained on massive datasets.

Researchers tested models from OpenAI, Google, Meta, and Anthropic and found they could infer race, location, job, etc.

Scammers could potentially exploit this to harvest personal data, and tech companies may already be using it for targeted advertising.

Experts say it's unclear how to prevent chatbots from inferring private information, posing concerning questions about privacy.