Dynamic Safety Governance for Financial Stability: A Regulatory Framework for Flow-Capacity Monitoring and Contingent Readiness under Model Ambiguity

Abstract

Purpose — This paper develops a regulatory framework for connecting system-wide liquidity surveillance to structural prevention, preparedness and intervention when financial flows, stress-period capacity and behavioral responses are uncertain. Design/methodology/approach — The paper integrates established work on macroprudential externalities, liquidity amplification, financial networks, coordination-sensitive crises and model uncertainty with recent system-wide regulatory practice. It develops a conceptual supervisory architecture and uses recent financial episodes as mechanism illustrations rather than predictive validation. OpenAI ChatGPT (GPT-5.6 Sol) was used to translate the author's original unpublished Chinese material into English and to copy-edit author-originated text; the author verified the submitted text and sources. Findings — Acute systemic pressure is better characterized by time-constrained flows converging on shared, state-dependent capacities than by aggregate volume alone. Capacity should be measured and, where safely possible, demonstrated operationally before residual uncertainty is represented as ambiguity. Model uncertainty need not imply passivity. Originality — The framework links cross-mechanism flow-capacity surveillance, decision-making under partial identification and bounded contingent readiness within one supervisory process. It also treats controlled reversible readiness drills and prospective immutable decision records as mechanisms for calibrating executable capacity and distinguishing forward performance from retrospective reconstruction. Research limitations/implications — The framework has not yet been prospectively validated or tested through live supervisory drills. Practical implications — Regulators can map common dependencies, calibrate recognised capacity to what institutions are willing and able to demonstrate, and pre-specify activation and exit conditions.