OpenAI wants to reassure companies. Its famous chatbot ChatGPT stopped remembering information provided by professional users in order to improve its answers as well as its ability to write credible text. “Our customers really don’t want us to train our model on their data, so we’ve changed our plans,” OpenAI CEO Sam Altman told CNBC. The change was quietly made in March. But the devil is in the details: this protection only applies to companies that pay a subscription to use OpenAI technology to build their own software solutions.
Nothing changes for Internet users who use the public version of ChatGPT. The conversations they have with the robots, as well as their opinion on the answers provided, are used to continue improving the capabilities of the robot. This point, which is not explained to the Internet user, has caused ChatGPT to be temporarily banned in Italy by the local Cnil.
It also causes cold sweats in companies. While ChatGPT has not yet “spit out” trade secrets provided by risk-averse employees, any data typed into it, such as confidential meeting notes or protected computer code, de facto enters OpenAI’s systems. . This has led several large groups, such as Samsung or Goldman Sachs, to block access to the platform.