AI风险观测站：从年报AI披露看社会韧性
原标题：The AI Risk Observatory: What Can We Learn from AI Disclosures in Annual Reports About Societal Resilience?
AI 摘要
该研究提出 AI Risk Observatory，用可复现的两阶段 LLM 分类流程处理 1,362 家英国上市公司 2020–2025 年（含部分 2026 年）的 9,821 份年报，并用 474 条人工标注段落验证方法。结果显示：提及 AI 风险的公司比例从 2020 年 2.8% 升至 2025 年 41.2%，AI 采用披露从 13.8% 升至 45.2%，供应商提及集中于以微软为首的小部分厂商；AIM 市场及能源、数据基础设施等关键国家基础设施部门的 AI 风险披露明显滞后，危害披露几乎缺失。作者进一步分类披露实质性，发现 2025 年虽 41.2% 报告提及 AI 风险，但仅 4.3% 属于实质性披露。
正文节选
The AI Risk Observatory: What Can We Learn from AI Disclosures in Annual Reports About Societal Resilience? Abstract Societal resilience research relies on access to useful and actionable data, which motivates our main research question: Can annual reports, processed at scale with LLMs, provide a useful signal about how companies disclose their response to AI? We test this by applying a reproducible two-stage classification pipeline to 9,821 annual reports from 1,362 UK listed companies (2020–20