生成式AI与监管审查：来自美国证监会常见信函的证据

标题	生成式AI与监管审查：来自美国证监会常见信函的证据
日期和时间	2026年1月7日（周三） 10:30-12:00
地点	会议楼II，103会议室
主讲人	舒展教授波士顿学院
摘要	This paper studies the use of generative AI at the Securities and Exchange Commission (SEC), focusing on a key enforcement tool – comment letters. Using the AI-detection tool GPTZero, we find that comment letters with higher AI usage are less comprehensive, and this relation holds in panel regressions as well as in an instrumental variables (IV) analysis. Further analysis reveals that comment letters with higher AI usage are more likely to omit important issues, such as topics related to revenue recognition, indicating that AI is not simply cutting “fluff.” The negative association between AI usage and comment letter extensiveness is more pronounced when SEC staff are busier and for industries requiring specialized knowledge. However, we find that it is attenuated, and in some cases reversed, when review teams include staff who are experienced, senior, or have accounting background. Finally, we find that coverage of small firms increased following the release of ChatGPT in November 2022, highlighting a potential benefit of AI. Taken together, our results provide early evidence on how generative AI influences regulatory oversight.
主讲人简介	舒展(Susan Shu)教授本科毕业于University of Dubuque，并于University of Rochester获得硕士及博士学位。她是波士顿学院Ernst & Young冠名讲席教授、会计学正教授，目前担任会计学博士项目主任。她的研究领域涵盖公司信息披露、传统媒体与社交媒体、监管执法、证券诉讼、内幕交易以及审计师-客户关系的经济学分析。其研究成果已发表在Journal of Accounting and Economics, Journal of Financial Economics, Journal of Accounting Research, and The Accounting Review等顶尖学术期刊。2014年，舒教授凭借合著论文 “Do Managers Withhold Bad News?” 荣获美国会计学会颁发的“会计文献杰出贡献奖”。