: ChatGPT, Bard and Ernie - AI-powered chatbots are all over the news, praising their prowess and highlighting challenges they face. Big tech companies behind these creations seem to be racing for media coverage and market share in a market that had not even existed in late 2022. This AI media battle is not the first, of course, with tech companies regularly jousting in areas such as autonomous vehicles, generative content, image analysis and voice assistants. What are the technical and legal risks associated with these technologies and the new uses they are opening up ? This article, the first in a series on generative AI, on the language processing and learning methods used to operate a chatbot, using ChatGPT as an
: From writing instructions on how to make a homemade bomb to refusing to provide a list of « bad » gifts to give, ChatGPT, which is being introduced as the representative of the new generation of chatbots, has made its choice. In this series of articles devoted to the use of generative AI through various experiments, we will first review some general information about these algorithms, then focus on the limitations encountered when using
: Before examining the means available to website publishers to block the collection of their data, it is first necessary to define the practices implemented by search engines and companies developing generative artificial intelligence models. Several concepts are used in this context: crawling , scraping , and TDM (Text and Data Mining). Although these terms refer to similar practices, their exact scope varies depending on usage and interpretation. It is therefore essential to clarify the meaning of each of these terms in order to establish a precise framework for the discussion.
: The emergence of agentic AI first raises the question of what this term actually means. Although the concept itself is not new, it has taken on a new significance in the era of generative AI. To date, there is no universally accepted definition of agentic AI or AI agents. This is illustrated by the OECD ’s February 2026 report, which identifies around ten different proposed definitions - each reflecting a distinct perspective. The terminology remains ambiguous, with varying scopes and overlapping concepts. Between marketing-driven labels, confusion with established terms such as conversational agent, and inconsistent definitions, it is far from straightforward to clearly define agentic AI and the systems it encompasses. In this article, we therefore provide an overview of both historical and contemporary terminology to help clarify the emerging—or perhaps not so new - landscape of agentic AI.