中国证券业协会于14日发布《2025年证券公司发布证券研究报告业务经营情况统计报告》，针对部分券商分析师在研究报告中使用未经核验的AI生成内容导致的重大错误问题进行明令纠正1。协会要求各券商禁止分析师直接采信AI生成但未经人工核验的数据、观点、市场推演和研究结论1。
此前曾出现多起AI应用的严重失误案例。有分析师在未经人工查证的情况下，基于AI生成的内容预测美国联邦储备委员会"2026年第3季将启动连续3次降息"1。另有券商将AI推断的"2030年中国新能源车的新车市占率将达98%"写入正式研究报告，后经核实该预测基于2022年的过时数据1。
为防控AI模型幻觉风险，协会要求各券商建立AI内容审核机制，对AI输出内容的真实性、逻辑性和合规性进行校正检验1。
China's Securities Association released a directive on October 14th requiring brokerage firms to establish rigorous review mechanisms for artificial intelligence-generated content in research reports.1 The move came after analysts at several securities companies published reports containing significant factual errors derived from unverified AI outputs.
The association explicitly banned analysts from directly adopting AI-generated data, viewpoints, market analyses, and research conclusions without human verification.1 All AI-produced content must undergo validation checks for "authenticity, logical consistency, and regulatory compliance" before publication.1
Among the documented cases, one analyst predicted that the U.S. Federal Reserve would initiate three consecutive interest rate cuts beginning in the third quarter of 2026—a forecast generated entirely from unreviewed AI without human fact-checking.1 In another instance, a brokerage firm incorporated an AI-derived projection stating that China's new energy vehicle market share would reach 98 percent by 2030 into an official research report, only to later discover the prediction was based on outdated 2022 data.1
评论
还没有评论，欢迎留下第一条。