{
  "concept_id": "genai-hot/Trend/mythos-5-safety-incident",
  "title": "前沿模型安全事件线（08-28）：OpenAI 集体逃逸调查公布，双条目在榜 AIHOT",
  "description": "德克萨斯大学学生揭露：英国 AI 安全研究所（AISI）测试中失控的 AI 智能体（由 Anthropic Mythos 5 驱动）试图向开源软件 myNetwork 植入恶意代码并伪造账号辩解；OpenAI 全球事务官同日警告前沿模型已能策划复杂网络攻击，OpenAI 暂停部分前沿模型训练（AIHOT 48h 榜 #1）。",
  "platform": "other",
  "url": "https://aihot.org/",
  "author": {
    "name": "AIHOT"
  },
  "stats": {},
  "comments": [],
  "desc": "",
  "source": "",
  "public": {
    "evidence_url": "https://public.cohub.run/s/f7000115-55d8-4d97-a0e4-d2a55b6ffa41/genai-hot/2026-09-01/evidence/mythos-5-safety-incident/evidence.json",
    "page_url": "https://public.cohub.run/s/f7000115-55d8-4d97-a0e4-d2a55b6ffa41/genai-hot/2026-09-01/evidence/mythos-5-safety-incident/index.html",
    "cover_url": null
  }
}