China calls for strengthening risk-based step-by-step management of the development and application of artificial intelligence in banks and insurance institutions, in order to better contribute to the real economy and meet the needs of the people, while effectively addressing the challenges posed by AI development.
The National Financial Supervisory Administration announced on Thursday that it has issued guidelines for the safe development and application of AI in the banking and insurance sectors.
Regulatory agency officials said that the guidelines not only aim to regulate the development and use of AI by banks and insurance institutions, but also to effectively prevent and control risks arising from the application of AI, promote the high-quality development of digital finance, promote the orderly integration of AI innovation and financial services, and guide the healthy and orderly development of the use of AI in the financial field in a beneficial, safe and fair manner.
The guidelines require financial institutions to strengthen top-level planning and overall governance by establishing a comprehensive AI lifecycle management framework and increasing oversight of application scenarios and business processes. Educational institutions are encouraged to build independently controllable, secure, and efficient intelligent computing infrastructure as needed. Large financial institutions with the necessary capabilities are encouraged to provide computing power services to smaller financial institutions and support industry efforts to jointly build and share infrastructure.
Additionally, financial institutions should incorporate AI-related risks into a comprehensive risk management framework, implement risk-based classification and tiered controls, and establish access controls for high-risk AI applications. Human oversight and intervention mechanisms must be implemented at key stages of high-risk applications, while outsourcing and supply chain risk management must also be strengthened.
The regulator said financial institutions should conduct regular assessments and reviews of AI-related risks and risk controls to prevent risks such as “black box” models (AI systems whose inner workings are invisible to users), AI hallucinations, and algorithmic bias, while strengthening cybersecurity, data security, and customer information protection.
Financial institutions must balance risk management and business development while building AI capabilities with security, transparency, and accountability. Data security and privacy protection must be strengthened, data classification and protection requirements strictly enforced, and content filtering and data desensitization measures improved, the official said.
Dong Shimiao, chief economist at Merchants Union Consumer Finance and executive director of Shanghai Financial Development Research Institute, said the guidelines represent the National Financial Supervisory Administration’s first dedicated regulatory framework for the safe development and use of artificial intelligence in the banking and insurance sectors.
He said the guidelines address the challenge of some financial institutions blindly implementing AI without proper regulatory guidance, while also establishing rules, defining red lines and setting the tone for AI applications in the banking and insurance sectors.
The guidelines go beyond risk management, emphasizing safe, practical and autonomous technology development and charting a path for high-quality AI development in the financial sector. These are designed to help AI contribute to the real economy, promote technological innovation and the orderly integration of financial services, and support the sustainable growth of digital finance, Dong said.
He said one of the notable features of the guidelines is that for the first time, high-risk application scenarios, such as fund trading, credit approval, underwriting and insurance claims settlement, are clearly defined. AI applications in these areas must be approved by the financial institution’s risk management committee and reported to regulators.
The guidelines also support co-building and sharing of computing resources across the industry, which can alleviate computing bottlenecks faced by smaller institutions and foster more balanced sector-wide development, Dong said.
Additionally, he noted that the guidelines set strong boundaries for protecting data privacy by explicitly prohibiting the use of sensitive personal information, such as names and identification numbers, in training and optimizing generative AI models.
Additionally, financial institutions need to maintain end-to-end oversight of their AI applications while continually improving model transparency. This will ensure compliance and traceability of algorithmic decision-making and provide an institutional framework for safe application of AI and healthy innovation in the financial sector, Dong said.
jiangxueqing@chinadaily.com.cn
