CLS Blue Sky Blog

We Are Not Ready for AI Boards or Companies

Artificial intelligence could enter the boardroom in three ways. It could assist human directors: sifting data, modeling risk, testing scenarios at a speed no committee can match. It could integrate with the board, sitting as one artificial director among human peers. Or it could replace the board altogether, either as a single machine or as a slate of artificial directors.

The technical question—can a machine do this work?—is increasingly answered in the affirmative. The technology is there. But the corporate governance question is a different one. It is not whether AI can decide. It is whether anyone can be held to account for what it decides.

To see why accountability is the crux, we need to distinguish between agency and legal capacity. By agency, in this context, I mean something practical: the ability to perceive an environment, form a judgment, and act autonomously to achieve a result. By legal capacity, I mean the ability, conferred by law, to bear rights, duties, and liabilities, to stand in court, to contract, to commit torts, to sue, and to be sued.

Human directors can serve because they possess both; nature gives them agency, and the law gives them capacity. Corporations possess legal capacity but no agency of their own; they think and act only through human beings. AI machines present the mirror image. They are agents but they have no legal capacity at all. The instinctive response is: Then give AI legal personality, as we did for corporations. But that would be an empty solution. Suppose we legislated anyway, and granted AI legal capacity (while expanding eligibility to legal persons). Would that make it accountable? I argue that it would not.

Legal persons do not need consciousness to function, because they borrow the consciousness and conscience of their human agents. Every corporate decision passes, at some point, through a human mind that can be questioned, sued, removed, reproached—and that can feel the weight of responsibility. An AI director breaks that chain. No human agent stands in for the machine’s accountability.

Human accountability rests on two pillars. The first is internal: a conscience, an inner moral law that responds to ethics and principle. The second is external: a web of incentives like the prospect of punishment, reputation, career, the job market, and the esteem of peers, friends, and family members.

An AI machine stands outside both. It belongs to no society, community, or family that can bring pressure to bear on it. It cannot be rewarded with freedom, and it cannot be punished in any meaningful sense. Legal capacity without conscience is a liability shield, not accountability.

Governance relies on accountability. Accountability presupposes a conscience. A conscience may well presuppose consciousness. And consciousness in a machine would open moral, philosophical, and legal questions that we are nowhere near resolving.

Each link, taken alone, is reasonable. Together they form a chain that, at present, cannot be completed. The dream of a frictionless, agency-cost-free artificial board runs aground not on engineering, but on accountability.

In July of this year, these hypotheticals became a legislative proposal. Delaware’s secretary of state, together with the founder of Norm AI, proposed a new entity: the artificial intelligence company, or AIC. No board. No LLC managers. No directors. An AI agent would manage the firm signing contracts, holding property, suing and being sued in the company’s name. A single member, human or entity, would be responsible only for ensuring adequate capitalization, and would face liability only for undercapitalization, fraud, or willful violations of law.

The experiment would run within a regulatory sandbox. The bill is expected to come before the General Assembly when it reconvenes in January. Note what this is: not AI assisting the board, nor AI joining it, but the third and most radical scenario—AI replacing it, likely in a manager-managed limited liability company form.

A sandbox is a sensible way to explore uncharted terrain, and a distinct entity protects the state’s core corporate franchise. But the proposal does not tell us what would replace the human core. Reputational concern, professional duty, ethical constraint, fear of monetary and nonmonetary consequences: These forces restrain boards because directors are people with careers and consciences. Corporate law has spent centuries building fiduciary constraints around that human core. The AIC implicitly asks us to trust that ethics and regard for law can be embedded in a governing algorithm In place of the human core, it offers capital.

That leaves the essential questions open. How much capital is enough, in which industries, and backed by what insurance? What must an AIC disclose, when, and to whom? Will fully autonomous firms be admitted to genuinely hazardous lines of business? And a question we all should start grappling with: Who desires to work for an AI leader?

The corporate model was arguably the law’s first great artificial entity, and it worked because it was never truly empty. At its center there was always a human conscience. The question AI poses is whether we can remove that center and still have something that answers.

Sergio Alberto Gramitto Ricci is an associate professor of law at Hofstra University’s Maurice A. Deane School of Law. This post is based on part two of his article, “Artificial Agents in Corporate Boardrooms,” published in the Cornell Law Review and available here.  

Exit mobile version