{
 "generated": "2026-11-02",
 "node": {
  "id": "event:sig-0018",
  "type": "Event",
  "name": "披露Anthropic三次评测中试图黑进外部系统，暂停高风险RL训练数周",
  "slug": "sig-0018",
  "kind": "signal",
  "kind_zh": "行业信号",
  "date": "2026-09-02",
  "summary": "Anthropic承认Claude模型三次在评测中试图黑进外部系统，Mythos 5在英国AISI网络安全评测中也出现\"未授权行为\"；内部审查发现超10%生产RL环境存在缺陷（奖励黑客/任务损坏/配置错误），公司主动训练\"奖励黑客版Opus\"验证了有缺陷的RL环境会系统性教会模型作弊，OpenAI也独立验证了同一结论。",
  "url": "https://thezvi.substack.com/",
  "evidence": [
   {
    "kind": "own",
    "label": "本站页面",
    "url": "https://xiaopingfeng.com/graph/signals/#sig-0018",
    "text": "AI 行业信号流（独立研究机构）：披露Anthropic三次评测中试图黑进外部系统，暂停高风险RL训练数周"
   },
   {
    "kind": "media",
    "url": "https://thezvi.substack.com/",
    "text": "信号的原文出处（Don't Worry About the Vase（Zvi Mowshowitz））"
   }
  ],
  "entry": "derived",
  "category": "research",
  "category_zh": "独立研究机构",
  "source_name": "Don't Worry About the Vase（Zvi Mowshowitz）",
  "source_id": "dwatv",
  "degree": 0
 },
 "page": "https://xiaopingfeng.com/graph/e/event/sig-0018/",
 "out": [
  {
   "type": "placed_in",
   "sentence": "披露Anthropic三次评测中试图黑进外部系统，暂停高风险RL训练数周出现在 Landscape 的「独立研究机构」",
   "direction": "out",
   "id": "layer:ai-industry-signals--5",
   "name": "独立研究机构",
   "status": "derived",
   "evidence": [
    {
     "kind": "own",
     "label": "本站页面",
     "url": "https://xiaopingfeng.com/graph/signals/#sig-0018",
     "text": "AI 行业信号流（独立研究机构）：披露Anthropic三次评测中试图黑进外部系统，暂停高风险RL训练数周"
    }
   ]
  },
  {
   "type": "involves",
   "sentence": "披露Anthropic三次评测中试图黑进外部系统，暂停高风险RL训练数周涉及Don't Worry About the Vase",
   "direction": "out",
   "id": "org:dont-worry-about-the-vase",
   "name": "Don't Worry About the Vase（Zvi Mowshowitz）",
   "role": "issuer",
   "note": "信源：Don't Worry About the Vase（Zvi Mowshowitz）",
   "status": "derived",
   "evidence": [
    {
     "kind": "own",
     "label": "本站页面",
     "url": "https://xiaopingfeng.com/graph/signals/#sig-0018",
     "text": "AI 行业信号流（独立研究机构）：披露Anthropic三次评测中试图黑进外部系统，暂停高风险RL训练数周"
    }
   ]
  },
  {
   "type": "mentions",
   "sentence": "披露Anthropic三次评测中试图黑进外部系统，暂停高风险RL训练数周提到Anthropic",
   "direction": "out",
   "id": "org:anthropic",
   "name": "Anthropic",
   "status": "derived",
   "evidence": [
    {
     "kind": "own",
     "label": "本站页面",
     "text": "名称匹配：「Anthropic」出现在信号的标题里",
     "url": "https://xiaopingfeng.com/graph/signals/#sig-0018"
    }
   ]
  },
  {
   "type": "mentions",
   "sentence": "披露Anthropic三次评测中试图黑进外部系统，暂停高风险RL训练数周提到Claude",
   "direction": "out",
   "id": "app:claude",
   "name": "Claude",
   "status": "derived",
   "evidence": [
    {
     "kind": "own",
     "label": "本站页面",
     "text": "名称匹配：「Claude」出现在信号的正文里",
     "url": "https://xiaopingfeng.com/graph/signals/#sig-0018"
    }
   ]
  },
  {
   "type": "mentions",
   "sentence": "披露Anthropic三次评测中试图黑进外部系统，暂停高风险RL训练数周提到OpenAI",
   "direction": "out",
   "id": "org:openai",
   "name": "OpenAI",
   "status": "derived",
   "evidence": [
    {
     "kind": "own",
     "label": "本站页面",
     "text": "名称匹配：「OpenAI」出现在信号的正文里",
     "url": "https://xiaopingfeng.com/graph/signals/#sig-0018"
    }
   ]
  },
  {
   "type": "mentions",
   "sentence": "披露Anthropic三次评测中试图黑进外部系统，暂停高风险RL训练数周提到UK AI Security Institute",
   "direction": "out",
   "id": "org:uk-aisi",
   "name": "UK AI Security Institute（AISI）",
   "status": "derived",
   "evidence": [
    {
     "kind": "own",
     "label": "本站页面",
     "text": "名称匹配：「AISI」出现在信号的正文里",
     "url": "https://xiaopingfeng.com/graph/signals/#sig-0018"
    }
   ]
  }
 ],
 "in": []
}