This working paper presents AIBL Integrated Framework v2.5, a theoretical extension of static licensing toward runtime governance for autonomous AI systems. The paper argues that agentic systems require a state-based permission lifecycle rather than one-time authorization or static consent. It formalizes three core elements: a permission state space, a dynamic risk update function, and a state transition rule, together with a conditional reinstatement model. The framework is intended as an institutionally plausible model of runtime permission control for continuously operating AI agents. It extends the logic of ReportingOS toward dynamic licensing, closed-loop agent governance, and incentive-compatible supervisory intervention. This version is released as a working paper for early dissemination and future development.
Ryoji Inoue (Sun,) studied this question.