OpenAI is developing an “incognito mode” for its popular chatbot ChatGPT that does not record users’ interaction history or use it to improve its artificial intelligence.
The San Francisco-based business has announced plans for a “ChatGPT Business” subscription with enhanced data restrictions.
The move comes as questions have been raised about how ChatGPT and other chatbots inspired by it manage hundreds of millions of users’ data, which is often used to improve or “train” AI.
“We’ll be moving more and more in this direction of prioritizing user privacy,” OpenAI’s chief technical officer Mira Murati said, with the goal of “it’s completely eyes off and the models are super aligned: they do the things that you want to do”.
She claimed that user data has helped OpenAI make its algorithms more dependable and decrease political bias, among other things, but that the company still faces hurdles.
With the product release on Tuesday, users can disable “Chat History & Training” in their settings and export their data.
OpenAI product executive Nicholas Turley, who compared it to an internet browser’s incognito mode, said the company would still keep discussions for 30 days to check for abuse before permanently deleting them.
Furthermore, the company’s upcoming business subscription will not use talks for AI model training by default.