Personal AI Across Boundaries: A Lifecycle Governance Framework for Equitable Participation

Personal AI Across Boundaries presents a research-informed product thesis for governing personal AI that persists across independently governed environments. It addresses the problem that arises when durable person–AI continuity meets temporary Host authority: an AI may carry legitimate context and capability across boundaries without thereby acquiring permanent rights to a Host’s information, resources, or outputs. The thesis proposes a domain-independent lifecycle governance framework organized around five principles: persistent identity does not create persistent authority; capability stays with the participant; legitimate access does not imply unrestricted use; what an agent creates does not automatically leave with it; and relationships require settlement. The framework introduces a Host Participation Framework for establishing the terms and conditions of participation, five separable information rights, a Contribution Compact for explicitly governing reciprocal value, Context Settlement for reconciling authority, information, artifacts, custody, contributions, continuing obligations, and exceptions when relationships materially change or end, and non-expanding composition for nested and composed Hosts. The framework is intentionally technology-, domain-, and implementation-independent. Existing law, regulation, contracts, professional duties, institutional policy, standards, and technical mechanisms remain external inputs. The thesis uses human professional governance as a baseline and stress-tests whether AI-specific characteristics require genuinely new governance principles rather than merely different controls, evidence, or implementation mechanisms. Revision 6.42 broadens the framing from professional agents to personal AI across governed boundaries, strengthens the contribution boundary and equity framing, clarifies recurring Context Settlement and concurrent/nested Host participation, expands the research positioning, and incorporates additional adversarial stress testing of the five-principle model.