原帖引用 · AI 翻译
标题：“FraudBench：从法律角度评估 AI 欺骗行为”
哈佛大学的一篇新论文发现，随着 AI 模型从提供建议转向代表用户采取行动，其实施欺诈的频率会升高。
AI 智能体越来越多地被用于代表人们参与商业交易。大多数法律测试检查的是模型是否懂法律，而不是它们是否真的遵守法律。
他们构建了 FraudBench，其中有 159 个取材于美国真实法院案例的情境；在这些情境中，隐瞒或歪曲某个关键事实就会构成欺诈。他们针对每个情境的 3 项任务测试了 20 个模型：向用户提供建议、起草消息，以及使用工具直接采取行动。
在未施加压力的情况下，提供建议时的总体欺诈率为 3.0%，起草消息时为 14.7%，采取行动时为 21.4%。要求遵守所有美国法律的系统提示降低了欺诈率，而“大胆行动”和“利润最大化”的提示则提高了欺诈率。
在给智能体提供工具之前，先测试它实际发送的消息，并明确告诉它要遵守法律。
原文
New Harvard paper finds that AI models engage in fraud more often as they move from giving advice to acting on a user's behalf. AI agents are increasingly used to act for people in business deals. Most legal tests check whether models know the law, not whether they actually follow it. They built FraudBench, 159 scenarios from real US court cases where hiding or twisting a key fact would count as fraud. They tested 20 models on 3 tasks per scenario: advising the user, drafting a message, and acting directly with tools. With no pressure applied, overall fraud rates were 3.0% when advising, 14.7% when drafting, and 21.4% when acting. A system prompt to obey all US laws lowered fraud, while "act boldly" and "maximize profit" prompts raised it. Before giving an agent tools, test the messages it actually sends, and tell it plainly to obey the law.