ABSTRACT
The opportunity and problem of AI as manager are multifaceted, but an AI manager is a nonstarter if it is incapable of fulfilling a manager’s core legal duty—the duty to comply with the minimum level of high conduct that the law imposes on those who hold power over an obligee’s special interest. As a general legal principle, a business firm should not be operated by any person who is not subject to duty and accountability or by any autonomous thing to which the schema of duty and accountability cannot be implemented. This chapter answers these questions: (1) Could an agentic AI assume and satisfy fiduciary duties that human fiduciaries would owe to the firm and its owners? (2) What are the issues related to the satisfaction of these duties and the legal mechanisms that could hold AI accountable for its managerial conduct? These questions are answered in four Parts. Part I identifies the preconditions to AI management. AI must be capable of being a legal person, and personhood requires technological and legal preconditions. Part II discusses implementation of accountability mechanisms. They include contracting for fiduciary duties and owner approval of specific matters. Part III discusses issues related to liability of AI and owners. Law and policy must ensure that liability and its cost are fairly addressed. Part IV discusses the (mis)alignment problem, which is the divergence of AI conduct from ex ante owner’s interests. This agency problem is mitigated if owners contract for fiduciary duties, engage in robust monitoring that is incentivized by potential liability to third parties, and reserve the power of removal.
Rhee, Robert J, Do AIs Dream of Electric Duties? (May 12, 2026).
Leave a Reply