Understanding confidentiality problems in Chatgpt: an approach based on data with the modeling of the LDA subject

This study examines the confidentiality problems associated with chatgpt, a prominent generative AI model, thanks to a data -based approach combining Twitter data analysis and a user survey. Taking the extraction of modeling techniques for modeling and categorization of data from the Latent Direct Allocation (LDA), research identifies the main areas of concern: 1) The confidentiality leak due to the exploitation of public data, 2) a confidentiality leak due to the exploitation of personal entries and 3) a confidentiality leak due to unauthorized access. The analysis of Twitter data of more than 500,000 tweets, supplemented by a survey of 67 Chatgpt users, reveals nuanced perceptions and experiences of users concerning the risk of confidentiality. A Python program was used to improve a 500K tweet data set referring to “Chatgpt” during the data preparation phase. To obtain a refined collection of terms, the steps included the conversion of the text into tiny, the elimination of mentions and hyperlinks, tokennized, the elimination of stopwords and the correspondence of keywords to extract tweets on the confidentiality of Chatgpt. Once the pre -treatment is over, there were 11,000 refined tweets. The results highlight significant apprehensions, in particular with regard to unauthorized access, stressing the importance of robust confidentiality measures in AI systems. The study contributes to understanding the concerns of users, clarifying political decisions and guiding future research on private life in generative AI. These studies could improve the safety and confidentiality of other AI systems and other systems. The public, companies, researchers, legislators and AI developers can all benefit from the useful information it provides to better understand and manage confidentiality.