It has been reported that the Italian privacy regulator ordered a ban on ChatGPT on Friday.
Prof Nello Cristianini, Professor of Artificial Intelligence at the University of Bath, said:
“Prompted by a data breach on March 20th involving the contents of user sessions and payment details, the Italian Privacy Authority has looked into chatGPT and today has ordered its urgent suspension in the national territory of Italy. The reasons listed for this decision go beyond the mere user session data, but include also the need for an age verification (to avoid exposing young children to potentially inappropriate content) and the issue of the data used to train the model itself, which is sourced from the web and might contain personal data.
“The most important considerations, for me as an AI researcher, are not so much about the user session data being inadequately protected, which prompted the investigation, or the lack of age-verification: both could be solved with routine solutions. Really concerning are instead these two: the training that may include also personal data without consent (without legal basis, the Authority says), and the possibility raised that these personal data might be treated in inexact manner, since the answers of chatGPT are not always correct. It is not clear how these can be fixed anytime soon.
“This goes to the heart of current technology, which needs to be trained on nearly-all of the web, and this will likely include also personal data, and which cannot guarantee correct answers. Furthermore recent studies do show that it is sometimes possible to “attack” a model of the same algorithmic family as GPT, to extract some of the training data.
“While it is not clear how enforceable these decisions will be, the very fact that there seems to be a mismatch between the technological reality on the ground, and the legal frameworks of Europe, seems to show that perhaps there was something to the letter signed by various AI entrepreneurs and researchers two days ago, calling for a 6 month ban on the deployment of AI technologies, to allow for our cultural tools to catch up.”
Declared interests
Prof Nello Cristianini: “the author of “The Shortcut – why intelligent machines do not think like us” edited by CRC Press.”