China's financial industry now has its first full provincial-level blueprint for embracing large models. On October 22, the Shanghai Financial Regulatory Bureau published “Measures on Promoting High-Quality Development of Digital Finance in Shanghai's Banking and Insurance Sectors”, rolling out 17 measures across five areas to deepen the fusion of digital technology and financial business in support of Shanghai's international financial center.
1. Five areas, 17 measures, evaluation first
The measures cover governance, service development, technology innovation, data resources and risk control: ① a digital finance evaluation system will be explored to monitor development levels and outcomes; ② inadequate policy implementation will be corrected; ③ risks and violations draw risk alerts, regulatory interviews, regulatory notices and administrative penalties.
2. AI scenarios scale up across the board
Technology application is the most concrete part: ① deepen smart marketing, smart credit, smart claims, smart risk control and internal management AI scenarios; ② optimize intelligent compute allocation, coordinate vertical model development, and build enterprise-grade AI platforms; ③ in tech finance, banks are encouraged to build intelligent approval and credit models for tech SMEs to shorten approval cycles; ④ supply-chain finance will use big data and blockchain to standardize receivables financing and supply-chain note financing products.
3. Compute and risk control: prudence is the keyword
For smaller banks and insurers, the document prescribes a pay-as-you-grow compute path: ① assess intelligent compute needs ahead of time under security and compliance, and obtain compute via the “rent-plus-buy” model based on national compute nodes or industry infrastructure, avoiding duplicated construction; ② peers are encouraged to jointly build and share digital infrastructure, with capable institutions opening data-center compute and technology platforms to the industry.
Risk control also gets a seat for AI: pilot vertical LLMs and general agents in financial risk control with security validation; explore cross-institution risk-information attestation and early-warning on blockchain infrastructure; and use AI to power security operations with intelligent threat monitoring and automated response. On data, the measures push asset inventory, classification and “smart data Q&A” capabilities.
Digitalization is a required course, but the path must match one's own scale—rent-plus-buy and shared building is the most pragmatic reminder for smaller institutions.
4. What to watch as the policy lands
Finance is among the most active arenas for industry LLMs, and Shanghai's 17 measures spell out “what to use, how to build, and how to govern” in one document: a scenario list for direction, an evaluation system as the yardstick, and security validation as the floor. For institutions driving financial digitalization, vertical model selection and private deployment capability will mark the next divide. NineZenith's Zenith-Act industry LLM supports private deployment for financial scenarios, with Zenith-Safety providing data and AI security compliance as modular support.
(Sources: ThePaper, 21st Century Business Herald, China News Service and other public reports)