Impact Evaluation on the European Privacy Laws governing generative-AI models -- Evidence in Relation between Internet Censorship and the Ban of ChatGPT in Italy

TL;DR

This paper evaluates how European Privacy Laws affected generative-AI models, specifically analyzing the impact of Italy's ChatGPT ban and internet censorship data using a hidden Markov model to understand user access limitations.

Contribution

It introduces a seven-state hidden Markov model to explain changes in internet censorship data related to the ChatGPT ban in Italy, linking legal regulations to user access impacts.

Findings

HTTP Invalid Requests decreased during the ban period

A seven-state hidden Markov model explains censorship data changes

EU regulations correlate with reduced internet access for users

Abstract

We proceed an impact evaluation on the European Privacy Laws governing generative-AI models, especially, focusing on the effects of the Ban of ChatGPT in Italy. We investigate on the causal relationship between Internet Censorship Data and the Ban of ChatGPT in Italy during the period from March 27, 2023 to April 11, 2023. We analyze the relation based on the hidden Markov model with Poisson emissions. We find out that the HTTP Invalid Requests, which decreased during those period, can be explained with seven-state model. Our findings shows the apparent inability for the users in the internet accesses as a result of EU regulations on the generative-AI.