# IA e diritto | Ferruccio Maria Sbarbaro | TEDxLink Campus University Source: https://www.youtube.com/watch?v=nAISkm7mHkw Recap page: https://rapidrecap.app/video/nAISkm7mHkw Generated: 2025-12-03T21:03:55.713+00:00 --- ## Quick Overview The feasibility of granting Artificial Intelligence systems legal personality, as discussed by Ferruccio Maria Sbarbaro, ultimately depends on whether existing legal frameworks, designed for human entities, can adequately address the challenges posed by AI's rapid, global evolution, particularly concerning liability and compensation for damages. **Key Points:** - The challenge of AI liability is complicated by the technology's rapid evolution, which often outpaces legal frameworks, demanding a flexible approach. - Existing legal tools like strict liability, vicarious liability, breach of duty of care, and product liability are insufficient because they rely on identifying a responsible human subject. - The existence of 'Emergent Capabilities' and 'Model Drift' in AI systems makes tracing responsibility back to a programmer or owner increasingly difficult. - The speaker cites the example of self-driving cars, noting that current systems might cause damages exceeding the $1 million capital/asset base often required for legal responsibility. - Current legal frameworks, particularly those based on the traditional paradigm of human will synthesis, struggle with AI's autonomous decision-making and cognitive capabilities. - Market solutions like Memberless LLCs and DAOs (Decentralized Autonomous Organizations) are emerging organizational forms that attempt to address AI's operational autonomy, independent of slow legal debate. - The core problem is establishing who pays for damages caused by autonomous AI, which requires either finding an accountable human or creating a new system of patrimonial responsibility for the AI itself. ![Screenshot at 00:45: The title slide introduces the topic 'Intelligenza artificiale e personalità giuridica: un incontro possibile?' \(Artificial intelligence and legal personality: a possible encounter?\) featuring the speaker Ferruccio Maria Sbarbaro and setting the stage for a legal discussion on AI.](https://ss.rapidrecap.app/screens/nAISkm7mHkw/00-00-45.png) **Context:** This TEDx talk, titled 'Lex Machina' and presented by Ferruccio Maria Sbarbaro at Link Campus University, explores the complex legal and ethical questions surrounding Artificial Intelligence (AI), specifically focusing on whether granting AI 'legal personality' is a viable solution for assigning liability when AI systems cause harm. The discussion contrasts traditional legal paradigms, rooted in human agency and will, with the characteristics of modern AI, such as autonomy, emergent capabilities, and distributed nature. ## Detailed Analysis Ferruccio Maria Sbarbaro argues that existing legal tools are ill-equipped to handle liability arising from advanced AI systems, which evolve faster and more globally than legislation. He emphasizes that the problem is concrete: when an AI causes damage (like in an autonomous vehicle incident), determining who pays is complex because current law requires attributing responsibility to a human subject (owner, programmer, etc.). This chain of responsibility becomes convoluted due to features like 'Model Drift' (AI changing its behavior post-deployment) and 'Emergent Capabilities' (unforeseen skills). Sbarbaro suggests that traditional legal concepts, which view legal organizations as syntheses of human will, fail when confronted with autonomous AI decisions. He contrasts the slow pace of the legal debate with the fast-paced market, which is already experimenting with new organizational structures like Memberless LLCs and DAOs to manage AI assets. He questions whether assigning a limited legal personality to AI, similar to that of a corporation, would be an effective remedial approach, noting that such a system must ensure an accountable patrimony exists to compensate victims, something that might not be guaranteed by current legal structures or by relying solely on human actors who might lack sufficient assets or foresight. The speaker concludes that the solution requires considering new legal tools that move beyond traditional human-centric liability models. ### Introduction to the Problem - The speaker introduces the complexity of AI liability, noting that the speed of technological and market evolution (like the IPO of ChatGPT) outpaces legal adjustments, forcing a reconsideration of existing legal principles. ### Limitations of Current Law - Existing legal concepts (strict liability, vicarious liability, product liability) are inadequate because they demand identifying a responsible human actor, which is difficult when AI exhibits autonomy and emergent capabilities. ### The Challenge of Attribution - Sbarbaro highlights that current law relies on attributing responsibility based on human will or custody, but autonomous, globally deployed AI systems challenge this framework, making it hard to assign fault or ensure compensation. ### Market vs. Law - While the legal debate is slow, the market is already adopting new organizational forms like Memberless LLCs and DAOs to manage AI entities, showing that practical solutions are developing ahead of legislation. ### Patrimonial Necessity - For any liability regime to be effective, it must guarantee a reachable patrimony (assets) to cover damages, a factor complicated by the unpredictable nature of AI risks and the difficulty of setting mandatory minimum capital requirements. ### Conclusion on Legal Personality - Granting AI a limited legal personality, similar to a corporation, is proposed as a potentially effective, albeit ideologically difficult, solution to ensure adequate remedial prospects for victims, even if the AI's 'will' is synthesized from human input. ![Screenshot at 00:04: The opening title slide 'LEX MACHINA' sets the theme of law and technology.](https://ss.rapidrecap.app/screens/nAISkm7mHkw/00-00-04.png) ![Screenshot at 00:45: A slide detailing the core questions of the talk: whether AI needs personality/subjectivity, and what tools might be used to achieve this.](https://ss.rapidrecap.app/screens/nAISkm7mHkw/00-00-45.png) ![Screenshot at 01:09: A slide summarizing the concrete problem: 'Il problema concreto: chi paga?' \(The concrete problem: who pays?\), outlining the need to adapt civil liability rules for AI.](https://ss.rapidrecap.app/screens/nAISkm7mHkw/00-01-09.png) ![Screenshot at 03:22: A slide defining an AI system based on the EU AI Act \(Art. 3\), emphasizing autonomy and adaptability, and showing its derivation from human design.](https://ss.rapidrecap.app/screens/nAISkm7mHkw/00-03-22.png) ![Screenshot at 05:01: A slide titled 'La cassetta degli attrezzi' \(The toolbox\) listing existing liability mechanisms like Vicarious Liability and Product Liability, noting their inherent difficulties in invoking the 'risk of development' defense.](https://ss.rapidrecap.app/screens/nAISkm7mHkw/00-05-01.png)