{"id":"arxiv/2607.23916","url":"https://arxiv.org/abs/2607.23916","tags":["math.NA"],"score":0,"title":"Recursive Governance: A Graph-Theoretic Framework for Risk Propagation and Drift Detection in Agentic AI Systems","domain":"math.NA","source":"arxiv","authors":["Sriram Nagaraj","Advaith Nila Narayanan"],"reviews":0,"subject":"math.NA","summary":"As financial institutions transition from traditional predictive models to autonomous agentic systems, the static model inventory requirements of traditional model risk management (MRM) face structural obsolescence. This paper proposes a dynamic Inventory-as-Code (IaC) governance loop that treats the model inventory as a living architectural component rather than a periodic documentation artifact. We make four principal contributions. First, we introduce a calibrated Degree of Autonomy (DoA) materiality score with an explicit, taxonomized tool-complexity weighting scheme that addresses the…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-07-27","reactions":{}}