OpenAI is launching a new formula around ChatGPT to attract businesses. This subscription includes settings intended to protect internal and sensitive information that the private sector might send to it.
This is one of the legitimate fears that companies may have about using a generative artificial intelligence tool like ChatGPT. What will happen to the data we send to AI? Are they preserved? Are they used for training the model? Can they be spat out to other people, at random from a prompt indiscreet person?
This would not be dramatic in the case of public information from an average company. But if we’re talking about internal, and potentially sensitive, information, that’s a different story. This problem arose very quickly, when ChatGPT became accessible to the general public. It became even more evident with the arrival of a tool for sending files.
These fears are not ignored by OpenAI. On the occasion of the presentation of a new type of subscription, ChatGPT Team, on Wednesday January 10, the American company planned a new setting dedicated to confidentiality. This will ensure that your discussions with the chatbot and the professional data you pass to it are not used to train its models.
“ You own and control your business data — we don’t learn about your data or conversations [passées via l’offre de ChatGPT Team, NDLR]and our models do not learn from your usage », Details the group in its announcement message. Added to this is a cascade of other benefits, detailed on the company website.
Are you interested in artificial intelligence news? Join our Artificial newsletter, which offers you a recap of the week’s topics on Numerama.
OpenAI provides tools and plans to exclude certain data from ChatGPT
This exclusion mechanism is not the first to be deployed by OpenAI. However, it is decisive for successfully conquering the business sector, and not just the general public. The professional environment constitutes a major growth lever for OpenAI, through the marketing of specific subscriptions: apart from ChatGPT Team, there is already ChatGPT Enterprise.
OpenAI has every interest in playing fair and scrupulously respecting what it says. The opposite would be quite simply disastrous in terms of reputation, if we learned that the data supposed to be excluded from the AI training field is in fact actually used to develop it. This would ruin its prospects and benefit the competition.
Launched at the end of August 2023, ChatGPT Enterprise has similar parameters. Therefore, the data from its customers is not used to train the models. It is also possible to define a personalized information retention period. Client prompts and the data they send are also not used to train OpenAI’s models.
Previously, in spring 2023, OpenAI deployed a more confidential chat mode. Here, the general public can request ChatGPT not to keep chat history. Deactivating this conservation prevents conversations from being used to train its artificial intelligence. However, you must know that this retention is taking place and cut it off.
Finally, there is a final parameter similar to the previous ones, but this time concerns access to online information. If ChatGPT can go online and take information, the chatbot is supposed to follow the site indexing rules. Clearly, it is possible to exclude certain pages or an entire site from ChatGPT. A crucial subject for the press.
Subscribe for free to Artificielles, our newsletter on AI, designed by AIs, verified by Numerama!