Research on the Cross-Domain Transmission Mechanism of Algorithm Ethical Risks in Enterprise AI Applications and the Construction of a Closed-Loop Governance System

Abstract

With the development of generative AI, machine learning-based decision-making and intelligent agents in marketing, risk control, recruitment, auditing and customer service have started spreading algorithmic ethical risks across departments, systems and suppliers as chain-like transmissions rather than as errors in single-point models. Based on the literature from the past three years on AI risk management, audit accountability, transparency and governance frameworks, this paper focuses on risk formation and diffusion to construct a governance model of "source identification - transmission measurement - tiered handling - continuous auditing - accountability loop". The research proposes that to enhance corporate AI governance, ethical values need to be transformed into explainable, accountable and correctable institutional mechanisms, and a closed-loop control system can be established through algorithm impact assessment, model cards, data genealogy, red team testing, third-party auditing and board-level governance.