https://hugo.jiahongw.com/tags/%E6%95%B0%E6%8D%AE%E6%B3%84%E9%9C%B2/ Recent content in 数据泄露 on ✏VictorHong's Blog Hugo -- gohugo.io zh 74349588616178697 68207800989737984 wujiahong2013@gmail.com (VictorHong) wujiahong2013@gmail.com (VictorHong) ©2021-2026 | VictorHong Sat, 28 Mar 2026 01:20:00 +0800 https://hugo.jiahongw.com/20260328/anthropic-claude-mythos-%E6%95%B0%E6%8D%AE%E6%B3%84%E9%9C%B2%E6%8F%AD%E7%A4%BA-ai-%E5%AE%89%E5%85%A8%E6%96%B0%E6%8C%91%E6%88%98/ Sat, 28 Mar 2026 01:20:00 +0800 wujiahong2013@gmail.com (VictorHong) Sat, 28 Mar 2026 01:20:00 +0800 https://hugo.jiahongw.com/20260328/anthropic-claude-mythos-%E6%95%B0%E6%8D%AE%E6%B3%84%E9%9C%B2%E6%8F%AD%E7%A4%BA-ai-%E5%AE%89%E5%85%A8%E6%96%B0%E6%8C%91%E6%88%98/ 核心观点 Anthropic 正在秘密测试一个名为 Claude Mythos 的 AI 模型，其内部代号为「Capybara」的新层级在网络安全能力上远超 Claude Opus 4.6，甚至让 Anthropic 自己都感到担忧— VictorHong featured image AI 网络安全 Anthropic Claude 数据泄露 大模型 技术