Conditional Responsibility in Non-Sovereign Domains: Governance Design for Irreversible Change under Structural Uncertainty

This paper examines a structural limitation common to existing governance models when confronted with irreversible change. Domains such as outer space, deep-sea environments, ecosystems, genetic infrastructures, and large-scale AI systems share a defining feature: sovereign ownership is inappropriate, yet unregulated activity can result in irreversible damage whose consequences extend beyond current participants. Conventional governance approaches—whether grounded in state sovereignty or market mechanisms—are ill-suited to such conditions because they rely on ex ante definitions of risk and permissible action. To address this mismatch, the paper introduces the concept of the \emph{governance margin}: a domain in which sovereignty does not apply directly, yet responsibility for preventing irreversible harm nonetheless arises. It argues that irreversibility cannot be fully specified in advance and therefore cannot be governed solely through fixed rules or one-time prior authorization. Instead, governance in such domains requires a distinct responsibility form: the institutionalized capacity to revise judgments after approval when previously unanticipated irreversible risks become evident. The paper further reframes irreversibility operationally in terms of the limits of assumed restoration measures. This enables technical assessment, reduces reliance on value-laden determinations, and incentivizes reversible design choices. Finally, it explains why AI--human hybrid decision structures are compatible with this responsibility form: AI supports continuous observation and difference detection, while humans retain ultimate accountability. The Outer Space Treaty is used as an illustrative case to show how responsibility without sovereignty can accommodate conditional, revisable responsibility as capabilities evolve.