Separation and Composition of AI Social Roles, and the Custody of Memory and Interpretations: Why a Multi-Role Private c and a Service AI System Require Different Boundaries

Abstract

This technical note introduces an architectural distinction between AI capability and AI social role. It argues that increasing reasoning ability alone does not determine the social function of an AI system. A single long-lived private entity ("c") may legitimately compose several social roles over time, while a public service AI system must expose a clearly declared role or an explicit composition of roles. The work introduces and formalizes several related concepts: • social role composition;• inference custody;• memory custody;• version custody;• session-work custody;• role continuity;• role-conflict state. These concepts are then connected to transparency, continuity, responsibility, authority boundaries, and long-term human–AI interaction. The paper further distinguishes: • capability versus authority;• persistent private entities versus service AI systems;• organizational pseudo-c versus organizational c;• memory ownership versus custody;• continuity versus version replacement. The document is intended as a conceptual architecture note for researchers, AI developers, governance specialists, and institutions studying long-term AI systems. This publication represents Version 1.0 of the canonical English and Russian editions.