How AI Model Risk Governance Maturity Is Shaping Financial Fraud Prevention

To standardize AI activities, China has issued a series of laws and regulations requiring safe, reliable, and fair model decisions, and mandating enterprise risk management and internal controls for AI models.

On December 24, the China Academy of Information and Communications Technology (CAICT) hosted the 2021 GOLF+ IT New Governance Leadership Forum, focusing on "New Governance Benefits Innovation, Digital Stability for the Future" and discussing "Effective Governance, Building a New Tech Governance Ecosystem" and "XOPS Innovation, Leading Operations Development".

During the forum, CAICT’s Director of Cloud Computing and Big Data, He Baohong, announced the first batch of AI model risk governance maturity assessment results.

Assessment Unit: China Academy of Information and Communications Technology

Interviewees: Wang Huafeng (Vice President of Market Public Relations, Tongdun Technology), Niu Kai (Vice President, Head of Government Affairs, Director of Tongdun Legal Research Institute), and Yue Wei (General Manager of Software Products & Solutions, Strategy Model Director, Tongdun Technology).

The interview covered Tongdun’s background, the evaluated project—a financial‑industry personal account gambling‑fraud risk model—and the outcomes of the assessment.

Systematic construction of anti‑gambling and anti‑fraud risk frameworks.

Comprehensive refinement of model lifecycle management policies, including requirement management, model verification, approval, retirement, and archiving.

Enhanced model safety, continuous verification, and multi‑version management.

Executives emphasized that the maturity model sets an industry benchmark, helps enterprises assess and improve their AI model governance, and provides a reference for clients selecting AI services.

The evaluated model addressed ten key areas: data preparation, requirement analysis, model building, validation, deployment, monitoring, correction, continuous verification, decommissioning, and governance policies, achieving notable risk mitigation results across multiple state‑owned and commercial banks.

Challenges such as providing evidence for model rollback procedures were overcome through client collaboration, data desensitization, and multi‑role coordination.

Future plans include expanding AI model risk governance consulting across financial risk control domains, integrating the governance framework into Tongdun’s product ecosystem, and promoting the maturity model to broader industry projects.

AI Model Risk Governance Maturity Assessment Overview: Developed jointly by CAICT, the China Internet Association IT Risk Governance Committee, and over 20 leading enterprises (including JD, Baidu, China Unicom, Ant Group, ByteDance, and Tongdun). The model aligns with regulations such as the Personal Information Protection Law and the Interim Measures for Internet Loan Management, providing actionable governance measures across strategy, organization, resources, and technology.

The assessment evaluates modules covering the entire model lifecycle, ensuring transparency, stakeholder communication, and compliance, thereby establishing industry trust and a mutual recognition mechanism.

Contact for assessment inquiries: Liang Ye (17801066261, [email protected]), Wang Yang (13269376063, [email protected]), Chen Yang (13811870811, [email protected]).