# Quando sbaglia l’AI | Alberta Antonucci | TEDxLink Campus University Source: https://www.youtube.com/watch?v=tXpBkHAz52k Recap page: https://rapidrecap.app/video/tXpBkHAz52k Generated: 2025-12-03T21:35:08.934+00:00 --- ## Quick Overview The central theme of the TEDx talk, delivered by Alberta Antonucci, revolves around the legal and ethical responsibility when Artificial Intelligence (AI) fails, asserting that the responsibility ultimately falls on humans—specifically those who use, fail to understand, or fail to govern the technology—rather than the AI itself, exemplified by the case of Samsung employees using ChatGPT for work assignments. **Key Points:** - The talk addresses the core question: When AI fails, who is responsible? The answer is that responsibility lies with the human user, the one who does not understand it, or the one who does not govern it, not the AI itself. - Antonucci cites a November 2022 incident where Samsung employees used ChatGPT to draft internal documents, leading to the company temporarily banning generative AI tools. - The speaker argues that the fundamental legal issue is not the AI's capability, but the human failure to govern it and the associated legal frameworks (like Italian Law 132/2025, mentioned as a future reference). - The speaker contrasts the 'no' response from ChatGPT when asked if it has consciousness with the fact that it still produces potentially harmful or false information that users trust implicitly. - The core problem identified is the human laziness or fatigue leading to the assignment of tasks to AI without verification, even when the AI produces legally significant documents or false data. - The ultimate conclusion is that humans must assume responsibility for the AI's output, emphasizing the need to educate students and professionals to critically evaluate AI-generated content. - The speaker uses the example of a professor in Texas asking students to create assignments using AI tools, leading to the work being judged as an algorithm's output rather than human effort. ![Screenshot at 02:15: Alberta Antonucci, wearing a bright yellow tweed suit, is introduced on stage, with the presentation slide displaying her talk title: "Quando sbaglia l’AI: le responsabilità che nessuno si vuole prendere" \(When AI errs: the responsibilities no one wants to take\).](https://ss.rapidrecap.app/screens/tXpBkHAz52k/00-02-15.png) **Context:** The presentation is part of a TEDx event titled 'Lex Machina' hosted at Link Campus University in Rome, focusing on the intersection of law and technology. Speaker Alberta Antonucci, a lawyer specializing in digital law and the web, explores the emerging legal and ethical dilemmas posed by generative AI tools like ChatGPT, particularly concerning accountability when these systems produce errors or misleading information in professional and academic settings. ## Detailed Analysis Alberta Antonucci opens her talk by recounting an anecdote where her daughter complained that 'Mom, today they blocked Word,' highlighting the rapid integration of technology into daily life, even for basic tasks. She quickly pivots to the main topic: the legal and ethical implications of AI errors, specifically referencing the case where Samsung banned generative AI after employees used ChatGPT to write internal code and documents in November 2022. Antonucci stresses that the issue is not the AI's inherent capability but the human failure to govern it, noting that while AI like ChatGPT cannot claim consciousness (as confirmed when asked directly), it generates outputs that people rely on without verification. She points to Italian Law 132/2025 as a future framework seeking to establish corporate policies regarding AI use. The speaker argues that the real difficulty lies in assigning responsibility: is it the user, the one who fails to understand the tool, or the one who fails to govern it? She suggests that the underlying problem is human laziness, where people trust the AI's output implicitly, even for critical tasks, leading to the company in the Samsung case effectively stopping work assigned to the AI. Antonucci concludes by urging that responsibility must remain human—not AI's—and emphasizes the need for education to cultivate critical thinking ('not to be a yes-man') so that users can verify the output, rather than simply accepting what the algorithm provides. ### Introduction and Context - The talk focuses on AI accountability (Lex Machina) - Speaker introduces the theme with an anecdote about her daughter and Microsoft Word - Cites the Samsung incident where ChatGPT was used for internal code/documents, leading to a temporary ban. ### The Core Problem - Responsibility Assignment: The central question is determining who is responsible when AI makes mistakes - The speaker asserts responsibility lies with the user, the non-understander, or the non-governor, not the AI itself - AI tools like ChatGPT do not claim consciousness. ### Legal and Ethical Frameworks - Mentions Italian Law 132/2025 as a future legal attempt to mandate corporate AI policies - Discusses the inherent human laziness that leads to over-reliance on AI for tasks like homework or legal drafting. ### The Danger of Unverified Output - AI can invent facts or produce misleading information that users trust implicitly - The speaker uses the example of a Texas professor whose students submitted AI-generated work, highlighting the lack of critical human oversight. ### Conclusion - The ultimate goal is to assume human responsibility: 'Not AI's, but our own' - Education must focus on teaching critical evaluation and verification skills, ensuring humans remain central to judgment and decision-making. ![Screenshot at 00:03: Title card displaying the event branding 'LEX MACHINA' across a black background.](https://ss.rapidrecap.app/screens/tXpBkHAz52k/00-00-03.png) ![Screenshot at 00:11: The speaker, an unidentified man, begins the opening remarks, welcoming the audience to TEDx Link Campus University.](https://ss.rapidrecap.app/screens/tXpBkHAz52k/00-00-11.png) ![Screenshot at 02:15: Slide introducing the speaker, Alberta Antonucci, and the talk title: 'Quando sbaglia l’AI: le responsabilità che nessuno si vuole prendere'.](https://ss.rapidrecap.app/screens/tXpBkHAz52k/00-02-15.png) ![Screenshot at 03:00: Alberta Antonucci on stage, gesturing while discussing how students use tools like ChatGPT for assignments.](https://ss.rapidrecap.app/screens/tXpBkHAz52k/00-03-00.png) ![Screenshot at 04:12: Slide showing an exterior view of a modern building with a large illuminated sign reading 'A&M Commerce', used as an example in the discussion about legal contexts in America.](https://ss.rapidrecap.app/screens/tXpBkHAz52k/00-04-12.png)