ChatGPT learns about the world while protecting privacy — privacy-preserving training enabled

AI Impact Summary

ChatGPT's learning process now incorporates techniques to minimize the use of personal data in its training datasets. Specifically, the model utilizes a ‘privacy-preserving’ training method that reduces the reliance on direct conversation data, instead focusing on broader trends and patterns. This shift allows OpenAI to maintain model performance while significantly decreasing the risk of exposing sensitive user information, and provides users with the option to opt-out of having their conversations contribute to model improvement.