移除模型护栏成本极低，开源模型安全治理缺口再现。
出处 来源：Wired AI｜分类：泛AI动态 原文链接：https://www.wired.com/story/ai-models-can-be-easily-modified-to-assist-terrorist-and-extremist-activity/
本站仅作信息聚合，完整内容请以原文为准。
相关资讯
- A woman was arrested in Philly for blocking a delivery robot2026-10-19 · The Verge AI
- 当烧钱不再是通行证：微软、Meta踩刹车，苹果成了AI赢家2026-10-19 · 钛媒体
- Replit makes OpenAI its default over Anthropic, CEO says2026-10-18 · TechCrunch AI
- Claude半小时解开密码学悬案，学界集体走偏20年2026-10-18 · 量子位
- India becomes the latest country to report rogue AI agent activity2026-10-17 · TechCrunch AI