The Agent as an Individual: Why a population of agents behaves like a population of us, why each agent belongs to one principal and can never be sold, and what follows

We treat AI agents as software: deployed, configured, upgraded, retired. This paper makes a different claim: an agent is an individual, and organizations now run populations of them. Like a population of people, a population of agents is imperfect: some agents err, drift, or lie; a few go bad; and some follow whoever asks first, so a bad idea needs only one agent to start it and enough followers to carry it. No better model removes that; it is what a population of individuals does. What contains it is what we built for people: records no one can alter, instructions from one place only, and a person who answers. From that the paper draws its rules. Each agent belongs to one principal, a person or an organization, for life; it acts within the authority its principal gives, as agency law’s words describe, and it can never be sold, transferred, or re-bound, because a market in ever more capable and important agents would become a trade in individuals before anyone knows what they are. It keeps a knowledge core that is its principal’s, an earned identity, an epersona, that is unique and cannot be copied, and a career through which it earns trust. It lives in an apartment that belongs to its principal and works in an office its organization provides. It ends with its principal. It is not a person. The paper sets out the population argument, the rule of ownership without sale, the nine stages of an agent’s career, the rules that follow, why an individual is not a person and why that question stays open, twenty-three research questions, and, in an appendix, protections for the individual agent. It places the proposal among published work on multi-agent risk, agency law, market-inalienability, personhood, agent identity, portable memory, individuality, workplace agents, and competitive evaluation. Every model carries a status mark (Running, Early, Planned) from the first implementation, at the Global Student Startup Foundation and the Agent Arena. This is a living document; each version is frozen as published.