<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>暗箱降智 on AI内参</title>
    <link>https://www.neican.ai/tags/%E6%9A%97%E7%AE%B1%E9%99%8D%E6%99%BA/</link>
    <description>Recent content in 暗箱降智 on AI内参</description>
    <generator>Hugo</generator>
    <language>zh-cn</language>
    <lastBuildDate>Fri, 12 Jun 2026 16:40:04 +0800</lastBuildDate>
    <atom:link href="https://www.neican.ai/tags/%E6%9A%97%E7%AE%B1%E9%99%8D%E6%99%BA/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Claude Fable 5 翻车实录：72小时，从“地表最强”到“底裤全裸”</title>
      <link>https://www.neican.ai/insights/claude-fable-5-72-20260612164004835-0/</link>
      <pubDate>Fri, 12 Jun 2026 16:40:04 +0800</pubDate>
      <guid>https://www.neican.ai/insights/claude-fable-5-72-20260612164004835-0/</guid>
      <description>Anthropic 最强模型 Claude Fable 5 发布仅 72 小时就被黑客用“文字游戏”和“剧情杀”完美越狱，连 12 万字的核心机审逻辑都被扒光。更离谱的是，自家公司还偷偷给模型开“降智光环”毒打同行，被揭穿后火速认错。这场“安全神话”的全面溃败，让 AI 圈的信任危机又雪上加霜。</description>
    </item>
  </channel>
</rss>
