Skip to content

GRA · 金融科技与转型

空间金融资本与危害、风险暴露和转型的交汇之处

人工智能与自动化

Statistical and machine-learning models now inform credit, underwriting, fraud detection, trading and operations. The Global Risks Alliance (GRA) researches the controls they need: validation, uncertainty, explainability, human oversight, third-party model management. It also asks the systemic question: what happens when many institutions rely on similar models and data at once? GRA builds, sells, audits and approves no model. Part of financial technology and transformation.

02模型的使用领域

Automation across the financial value chain.

信贷

贷款决策

Scoring, limit setting and early-warning models, where errors affect access to credit and fairness as well as losses.

保险

Underwriting and claims

Risk selection, pricing inputs and claims triage, where model drift after a changing climate or market can mislead.

运营

Fraud and operations

Transaction monitoring, document processing and customer service, where automation scales both efficiency and failure.

03规则

AI in finance: what applies, and when.

工具状态,2026 年 10 月
EU AI Act high-risk duties for creditworthiness and life and health insurance pricingApply from 2 December 2027 under the AI Omnibus (Regulation 2026/1744)
US revised model risk management guidance (SR 26-2)Issued April 2026; generative and agentic AI outside its scope
PRA SS1/23对获得模型批准的英国银行已生效
NIST 人工智能风险管理框架自愿性
04地球观测

Machine learning on satellite data, for exposure.

One growing use of AI in risk finance is extracting exposure and damage from imagery: building footprints, flood extent from Sentinel-1 radar, crop condition. These models change exposure layers quickly after an event and carry their own error rates, which must travel with the layer. Methods at GCRI disaster monitoring.

05专业控制措施

完善的模型治理有哪些要求。

  1. 01

    验证

    Independent testing of a model’s design, data and performance against its intended use, before and during deployment.

  2. 02
  3. 03

    可解释性

    Being able to state why a model produced an output, to the people it affects and to those who oversee it.

  4. 04

    人工监督

    设置明确的节点,使有能力的人员可以审查、否决或停止自动化决策。

  5. 05

    第三方

    Knowing what is inside a vendor model, how it changes and what happens if the vendor fails.

06系统视角

当所有人都使用同一模型时。

模型风险 通常由各机构分别管理。GRA 研究共同情形。共同的供应商、训练数据和方法可能使误差相关,导致许多机构在同一时刻误判同一风险。

  • 哪些模型和数据来源在整个行业中被广泛共享?
  • How would a common error show up in credit, claims or market behaviour?
  • 哪些信号能及早揭示相关性模型失效?
  • Which questions belong to supervisors, auditors or each institution’s own validation function?
07边界与连接

仅为研究,不构成保证。

GRA validates, audits, rates and certifies no model or vendor. Its research is not a compliance opinion. Each institution stays accountable for its own models. The risk channel: AI and model risk. A proposed AI and Financial Model Risk community is an established platform. Related lines: digital assets research and quantum and cryptographic transition. The Global Risks Forum (GRF) discusses AI in public decisions on digital and AI governance.

09问题

问题与解答。

GRA 在自身研究中使用 AI 吗?
Where GRA uses automated methods in research, the publication states the method and its limits, and findings are reviewed by a person before release.
Can GRA review my institution’s model?
No. GRA does not perform any institution’s own model validation or risk-management work. Its outputs are industry-wide research.
这与 GRA 平台 Fintech Nexus 有关吗?
Yes. Fintech Nexus covers AI in financial services as part of its sector community; this page frames the cross-cutting research questions.
How can practitioners contribute?
模型验证人员、数据科学家和风险管理人员可通过 会员申请表登记意向。工作组将在其条款通过后开放。
报告贡献

从本页贡献证据。

Reports 是联合体的季刊。您对该主题的了解可以成为下一期中的证据、案例或更正,署名归您,并在属于您本人作品时以 CC BY 4.0 发布。

全球风险联盟(Global Risks Alliance)
全球风险联盟(Global Risks Alliance)

向联盟提出风险暴露或资本问题。

信息和咨询现已开放。每项咨询都会获得一个编号,并告知下一步;申请、付款、算力分配和注册只有在相应团队、条款和审查能力就绪时才会开放。