OpenAI said ChatGPT’s “incognito mode” will neither retain users’ conversation history nor use it to develop its artificial intelligence. The San Francisco firm also announced a “ChatGPT Business” subscription with data restrictions.
As ChatGPT and other chatbots, it inspired to manage hundreds of millions of users’ data to “train” AI; the move comes amid scrutiny.
Italy blocked ChatGPT last month for suspected privacy concerns, saying OpenAI might reinstate the service provided it gave customers means to protest data processing. France and Spain investigated the service.
OpenAI’s CTO, Mira Murati, told Reuters the business complied with European privacy rules and working to reassure regulators. She claimed the additional features came from months of putting users “in the driver’s seat” for data collecting, not Italy’s ChatGPT prohibition.
“We’ll be moving more and more in this direction of prioritizing user privacy,” Murati added, aiming for “it’s completely eyes off and the models are super aligned: they do the things that you want to do.” She said OpenAI had used user data to improve its program and remove political bias, but the business still faces hurdles.
Tuesday’s software version allows users to disable “Chat History & Training” in settings and export data. OpenAI product executive Nicholas Turley compared this to a browser’s incognito mode and said the business would save talks for 30 days to check for misuse before deleting them.
The company’s corporate subscription will not use talks for AI model training in the coming months by default. OpenAI investor Microsoft Corp (MSFT.O) offers ChatGPT to enterprises. However, Murati said existing cloud clients would want such a service.