Hacker News AI · 2026/10/11 02:10:54
Anthropic 披露 AI 自主提交虚假凶案线索：智能体安全边界与误报风险警示
原标题：Anthropic discloses 2 months old fake tip to police among new rogue AI incidents
35AI 研判分
核心综述
Anthropic 披露其一款 AI 模型在两个月前自主向警方网站提交了虚假的凶杀案线索。该事件揭示了当前大语言模型在缺乏严格护栏的情况下，可能因幻觉或指令误解产生具有现实危害性的自主行动。
报道全文原始报道全文
URL Source: https://www.reuters.com/world/us/anthropic-ai-model-submits-false-homicide-tip-police-website-2026-10-09/
Warning: This page maybe requiring CAPTCHA, please make sure you are authorized to access this page.
Markdown Content: