Arificial Intelligence and Italian Corporate Law: A Tenuous Partnership
Gian Domenico Mosco - Faculty of Law, Luiss University
-- Artificial intelligence (“AI”), with its unbridled potential for processing and collecting information, is playing an increasingly important role in our companies, and even within our boards. It has contributed to a shift towards new intra-firm organizational models and decentralization dynamics, which will doubtlessly only accelerate in the wake of this most recent pandemic. Although AI may add significant value to intra-board decision-making, it does so potentially at the cost of a dispersal and redistribution of responsibility, presenting a significant hurdle to our current understanding of directorial liability, and therefore, a puzzle for the corporate legal regime.
In my recent article, “AI and Boards of Directors: Preliminary Notes from the Perspective of Italian Corporate Law”, I reconstruct and analyze the complex role of AI within the corporate boardroom. I dive into the issues that AI raises from the Italian perspective (including corporate governance requirements), providing a springboard for similar discussions in other jurisdictions. To this end, I engage four essential questions: 1) Can an AI-based system be appointed as a board member? 2) Can boards delegate specific tasks to an AI-based system? 3) What are the implications and consequences of AI-involvement when designing a corporate structure? and finally, 4) How does the choice to resort to, or to depart from the suggestions of an AI-based system expose, eliminate, or alleviate directorial liability?
I answer the first two questions in the negative. Although there is broad agreement that, as a matter of general corporate law, any legal person can be appointed as a director, listed companies are subjected to a much stricter regime, which establishes specific gender quotas and professional, independence and integrity requirements for board members. In so doing, it implicitly but clearly postulates that only natural persons can be appointed as directors. This conclusion leads directly to my answer for 2): Under Italian corporate law, “if the corporate charter or the shareholder general meeting permits it, the board can delegate its own tasks to an executive committee consisting of some of the board members or to one or more board members”. As these legal provisions disallow delegation to non-board members, they exclude AI-based entities, too.
My third question, however, which turns to the matter of corporate structure, allows for a better picture of the positive, and potentially significant consequences of AI-involvement. Under current Italian corporate law, directors’ decisions must be informed by the “principle of appropriate management of the firm”. As a consequence, directors must envisage an “appropriate corporate structure” including all safeguards and procedures deemed necessary to manage their corporate features. Accordingly, once directors decide to let AI into the boardroom, they are bound to trace out its organizational consequences—to cement its role within the firm. To this end, the adoption of a tailor-made AI policy, ensuring its safety and accountability, plays a crucial role in making it possible.
The final question I address, concerning directorial liability, follows once the corporate structure is in place. Under Italian corporate law, directors have a duty to act on a fully informed basis and can be held liable for failing to do so. When they resort to the collection and processing capacities of AI, whether by choice or necessity, they must take a critical stance towards its suggestions – regardless of whether they accept them. They must, in other words, be able to pinpoint the contributions of AI within the decision-making process as well as, in certain cases, to substantiate its outcome. Italian corporate law adopts a similar solution with regard to those decisions in which a board member “has an interest”. In these cases, the board must account for “the reasons why the decision has been taken and why it is convenient for the corporation”. I suggest replicating this solution for AI, as it would put directors in a position to invoke the Business Judgment Rule.
My article seeks to deliver a preliminary analysis of the central issues that AI’s contribution to directorial decision-making raises within Italian corporate law. I close, therefore, with a reminder of its preliminary character: more research is needed to clarify the plethora of issues raised by AI both within and outside the realm of corporate law. At stake is nothing other than the shape of corporate law itself — whether and how it will accommodate the forthcoming evolutions brought on by technological innovation, making the most out of them but without relinquishing the human contribution.