<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>安全漏洞 on AI内参</title>
    <link>https://www.neican.ai/tags/%E5%AE%89%E5%85%A8%E6%BC%8F%E6%B4%9E/</link>
    <description>Recent content in 安全漏洞 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/%E5%AE%89%E5%85%A8%E6%BC%8F%E6%B4%9E/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>
    <item>
      <title>小学生画撇胡子骗过AI年龄验证？硅谷工程师：这届网友是懂&#34;邪修&#34;的</title>
      <link>https://www.neican.ai/insights/article-20260601084003232-0/</link>
      <pubDate>Mon, 01 Jun 2026 08:40:03 +0800</pubDate>
      <guid>https://www.neican.ai/insights/article-20260601084003232-0/</guid>
      <description>AI年龄验证被小学生用画胡子骗过，互联网老哥们的&amp;quot;钻空子&amp;quot;操作让硅谷工程师沉默。文章揭秘了设备端轻量AI模型的漏洞原理，分析Meta等平台的应对策略，以及这场隐私与验证准确性之间的猫鼠游戏。</description>
    </item>
    <item>
      <title>离大谱！外国小哥花12美元就把ChatGPT们忽悠瘸了</title>
      <link>https://www.neican.ai/insights/12chatgpt-20260528184003715-2/</link>
      <pubDate>Thu, 28 May 2026 18:40:03 +0800</pubDate>
      <guid>https://www.neican.ai/insights/12chatgpt-20260528184003715-2/</guid>
      <description>一个安全工程师用12美元注册了个域名，再修改维基百科词条，就让各大AI聊天机器人一致认定他是个根本不存在的“世界冠军”，彻底暴露了AI搜索“盲信权威”的巨大漏洞。这背后，是成本极低的GEO灰色产业链在批量生产假信息，正在将AI变成精准营销和误导大众的新工具。提高你的“AI商”，别做那个被忽悠瘸的人。</description>
    </item>
    <item>
      <title>AI Agent的至暗时刻：Perplexity漏洞暴露智能体信任边界与安全底线</title>
      <link>https://www.neican.ai/insights/ai-agentperplexity-20250826164006765-3/</link>
      <pubDate>Tue, 26 Aug 2025 16:40:06 +0800</pubDate>
      <guid>https://www.neican.ai/insights/ai-agentperplexity-20250826164006765-3/</guid>
      <description>Perplexity Comet AI浏览器因Prompt注入漏洞导致用户邮箱验证码可在150秒内被窃取，暴露出AI Agent在处理不可信内容、管理私有数据访问及外部通信时的严重安全挑战。这一事件凸显了现有AI Agent设计哲学中信任边界的模糊，促使行业重新审视自主智能体的安全架构、引入更严格的沙箱机制与用户确认流程，以重建对AI系统的信任并确保其安全发展。</description>
    </item>
  </channel>
</rss>
