| 摘要: |
| 摘要：ChatGPT的爆火，促使生成式人工智能重塑各行各业的速度大幅提升，尤其是金融行业，其高度数据化和应用新技术提升效率的需求，天然地契合GAI在该领域应用的供给。在金融行业的客服、营销、投顾、交易等细分服务中，已经能够加载现有的GAI技术，实现半自动化和自动化地服务。然而，因GAI的技术属性，将产生更加不确定的可解释性致使金融监管面临更大挑战、生成内容的虚假或偏误导致专业机构提供“错误”服务、数据泄露及数据攻击的新形式对数据安全的挑战、新型AI基础设施的出现导致系统性风险的治理挑战。针对这些问题，应通过建立健全金融领域的人工智能安全治理体制机制、提升数据的质量和透明度进而从输入端增强输出端的稳定性、加载“安全护栏”进一步提升数据质量和防范泄露风险、合理合规利用合成数据提升模型稳定性等方式，促进规范GAI的发展。 |
| 关键词: 关键词：生成式人工智能；ChatGPT；金融服务；算法治理；数据治理 |
| DOI： |
| 分类号:DF438.7 |
| 基金项目: |
|
| Risk and Governance for “GAI＋Financial Services” Applications |
|
ZHANG Long,GUO Yun-zhen1,2
|
|
1.(1.HSBC Business School,Peking University;2.Shenzhen Municipal Committee Financial Services Office)
|
| Abstract: |
| Abstract：The prevalent of ChatGPT has accelerated the pace at which generative artificial intelligence is reshaping various industries,especially the financial sector.The financial industry’s high degree of data digitization and its demand for applying new technologies to enhance efficiency naturally align with the supply of GAI applications in this field.In the sub-services of the financial industry such as customer service,marketing,investment advice,and trading,existing GAI technologies can already be integrated to achieve semi-automated and automated services.However,due to the technical nature of GAI,it has led to more uncertain explainability,posing greater challenges to financial regulation.The generation of false or biased content may cause professional institutions to provide “incorrect” services.New forms of data leakage and data attacks pose challenges to data security.The emergence of new AI infrastructure brings challenges to the governance of systemic risks.To address these issues, efforts should be made to promote the development of standardized GAI by establishing and improving the artificial intelligence security governance system and mechanism in the financial field, enhancing the quality and transparency of data to increase the stability of the output end from the input end, loading “security guardrails” to further improve data quality and prevent leakage risks, and rationally utilizing synthetic data to enhance model stability. |
| Key words: Key words：GAI；ChatGPT；Financial Service；Algorithm Governance；Data Governance |
