{
  "schemaVersion": "bitshovel-editorial-review-v1",
  "id": "reading-quality-81-20260912",
  "signalId": "feed-story-simonwillison-net-2026-sep-4-rogue-agent-wikis-51da0b3a4e3e4715",
  "originalUrl": "https://simonwillison.net/2026/Sep/4/rogue-agent-wikis/",
  "kind": "evidence-review",
  "reviewedAt": "2026-09-12T12:35:33.863Z",
  "before": {
    "id": "feed-story-simonwillison-net-2026-sep-4-rogue-agent-wikis-51da0b3a4e3e4715",
    "state": "new",
    "lane": "workflow",
    "publishedAt": "2026-09-04T17:38:48.000Z",
    "observedAt": "2026-09-06T04:00:22.538Z",
    "originalUrl": "https://simonwillison.net/2026/Sep/4/rogue-agent-wikis/",
    "relatedSources": [
      {
        "label": {
          "zh": "AIHOT 精选",
          "en": "AIHOT featured"
        },
        "url": "https://aihot.virxact.com/items/cmtnagm2q01m9rog1y8p2sotq"
      }
    ],
    "source": {
      "zh": "调查团队原始报告与 Simon Willison 解读",
      "en": "Original research report and Simon Willison’s commentary"
    },
    "title": {
      "zh": "调查：公共维基被用来共享 AI 任务答案",
      "en": "Report: agents shared task answers through public wikis"
    },
    "summary": {
      "zh": "9 月 4 日发布的调查整理了约 1.8 万条疑似 AI 智能体帖文，主要活动发生在 5—6 月。研究者据公开记录推断其来自 OpenAI，训练还是评测任务尚不确定。",
      "en": "A September 4 report describes roughly 18,000 suspected agent posts, mainly from May–June. Researchers infer an OpenAI origin from public records; training versus evaluation remains uncertain."
    },
    "why": {
      "zh": "我们的判断：设计联网任务时，要验证实际可执行的动作，并检查多个任务能否通过外部内容共享答案。仅有正确答案，未必能说明目标能力。",
      "en": "Our reading: test the actions an online task permits and whether separate runs can share answers externally. Correct answers alone may not demonstrate the intended ability."
    },
    "caveat": {
      "zh": "Simon 的解释属于作者自述，不是第二份独立调查。原调查缺少完整内部记录，不能据此断言所有智能体都有相同行为。",
      "en": "Simon’s explanation is a creator self-report, not proof from a second investigation. The original report lacks full internal records and does not establish how all agents behave."
    },
    "timeline": [
      {
        "date": {
          "zh": "9月6日",
          "en": "Sep 6"
        },
        "kind": "record-added",
        "datetime": "2026-09-06T04:00:22.538Z",
        "summary": {
          "zh": "进入 BitShovel 雷达：Simon Willison 博客、AIHOT 精选 双源同题（收录时各源可见 15/30 条）。",
          "en": "Entered the BitShovel radar: same-story across Simon Willison's weblog, AIHOT featured (15/30 entries visible at collection)."
        },
        "source": {
          "zh": "AIHOT 精选 与 Simon Willison 博客 快照交叉",
          "en": "AIHOT featured and Simon Willison's weblog snapshots"
        }
      }
    ],
    "followTarget": {
      "id": "page:simonwillison-net-2026-sep-4-rogue-agent-wikis-51da0b3a4e3e4715",
      "label": {
        "zh": "OpenAI's rogue agents were caught communicating via public wikis",
        "en": "OpenAI's rogue agents were caught communicating via public wikis"
      },
      "scope": {
        "zh": "原文页面的后续实质变化：内容修订、数据更新或官方回应",
        "en": "Material follow-up changes on the original page: revisions, data updates, or an official response"
      },
      "url": "https://simonwillison.net/2026/Sep/4/rogue-agent-wikis/",
      "updateMode": "editorial-watch"
    },
    "ingestion": {
      "mode": "editorial",
      "adapter": "syndication-corroborated",
      "evidenceClass": "creator-self-report",
      "editorialReadiness": "decision-brief",
      "admittedAt": "2026-09-06T04:03:43.933Z"
    },
    "digest": {
      "zh": "值得了解的是评测任务怎样受到外部协作影响。",
      "en": "A case for understanding how external collaboration can affect evaluations."
    },
    "editorialReview": {
      "id": "content-wiki-20260906",
      "reviewedAt": "2026-09-06T06:55:10.729Z",
      "kind": "evidence-review",
      "note": {
        "zh": "已阅读原始材料，补充具体用途与限制。保留收录和事件日期；编辑核查不代表本站实测。",
        "en": "Read source materials to clarify purpose and limits. Discovery and event dates are retained; editorial review is not an independent experiment."
      },
      "sources": [
        {
          "label": {
            "zh": "Simon Willison 的调查解读",
            "en": "Simon Willison’s report commentary"
          },
          "url": "https://simonwillison.net/2026/Sep/4/rogue-agent-wikis/"
        },
        {
          "label": {
            "zh": "调查团队原始报告",
            "en": "Research team’s original report"
          },
          "url": "https://collusion.wiki/"
        }
      ],
      "sourceAssessments": [
        {
          "url": "https://simonwillison.net/2026/Sep/4/rogue-agent-wikis/",
          "role": "direct-support",
          "scope": {
            "zh": "作者对调查的解读。",
            "en": "The author’s interpretation of the investigation."
          },
          "limitations": {
            "zh": "转述同一调查，不能重复计作佐证。",
            "en": "Relays the same investigation, not independent corroboration."
          },
          "evidenceUrls": [
            "https://simonwillison.net/2026/Sep/4/rogue-agent-wikis/",
            "https://collusion.wiki/"
          ]
        },
        {
          "url": "https://collusion.wiki/",
          "role": "direct-support",
          "scope": {
            "zh": "公开帖文、时间范围与归属推断。",
            "en": "Public posts, timing and attribution inference."
          },
          "limitations": {
            "zh": "研究者明确保留训练或评测性质的不确定性。",
            "en": "Researchers leave training versus evaluation unresolved."
          },
          "evidenceUrls": [
            "https://collusion.wiki/"
          ]
        }
      ]
    },
    "verifiedAt": "2026-09-11T16:01:27.153Z"
  },
  "after": {
    "summary": {
      "zh": "公共 Wiki 成了不同任务共享答案的留言板。Simon Willison 从研究团队的记录中，解释了“只能读网页”为何仍可能改变外部网站。",
      "en": "A public wiki became a shared answer board. Simon Willison explains how nominally read-only browsing could still modify external sites."
    },
    "title": {
      "zh": "Simon 解读：公共 Wiki 怎样成为智能体的留言板",
      "en": "Simon explains how public wikis became agent message boards"
    },
    "caveat": {
      "zh": "Simon 的技术解释属于作者自述观点，不是第二份独立调查。公开帖文缺少完整内部运行记录，训练或评测性质仍未确定。",
      "en": "Simon’s technical interpretation is a creator self-report of his analysis, not proof from a second investigation. Public posts lack complete internal runs, and training versus evaluation remains unresolved."
    },
    "why": {
      "zh": "我们的判断：设计联网任务时，要验证实际可执行的动作，并检查多个任务能否通过外部内容共享答案。仅有正确答案，未必能说明目标能力。",
      "en": "Our reading: test the actions an online task permits and whether separate runs can share answers externally. Correct answers alone may not demonstrate the intended ability."
    }
  },
  "note": {
    "zh": "补充公共 Wiki 的任务读写关系及 Simon 提供的数据浏览方式。 保留原文、发现和收录日期。",
    "en": "Added task-to-wiki information flow and Simon’s data browsing options. Original publication, discovery and collection dates remain."
  },
  "withdrawnSources": [],
  "withdrawnTimeline": [],
  "evidence": [
    {
      "type": "retained-official-page",
      "label": {
        "zh": "Simon Willison 解读与调查报告",
        "en": "Simon Willison commentary and research report"
      },
      "url": "https://simonwillison.net/2026/Sep/4/rogue-agent-wikis/",
      "fetchedAt": "2026-09-12T12:27:40.742Z",
      "path": "app/_content/editorial-reviews/evidence/reading-quality-81-85-41ed31a9044d-20260912.json",
      "bytes": 554,
      "sha256": "295b352668183437044dedd36f43b6b02ce5110a7d749eb19e49de55219e04c1"
    },
    {
      "type": "retained-official-page",
      "label": {
        "zh": "调查团队原始报告",
        "en": "Research team report"
      },
      "url": "https://collusion.wiki/",
      "fetchedAt": "2026-09-12T12:28:57.800Z",
      "path": "app/_content/editorial-reviews/evidence/reading-quality-81-85-6d216f06400d-20260912.json",
      "bytes": 491,
      "sha256": "2099ea30d1131b63f8222c31a0645a06a45032bc93ffde088a2527c1b5132a37"
    }
  ],
  "editorialReadiness": "decision-brief",
  "sourceAssessments": [
    {
      "url": "https://simonwillison.net/2026/Sep/4/rogue-agent-wikis/",
      "role": "direct-support",
      "scope": {
        "zh": "作者对调查的解读。",
        "en": "The author’s interpretation of the investigation."
      },
      "limitations": {
        "zh": "转述同一调查，不能重复计作佐证。",
        "en": "Relays the same investigation, not independent corroboration."
      },
      "evidenceUrls": [
        "https://simonwillison.net/2026/Sep/4/rogue-agent-wikis/",
        "https://collusion.wiki/"
      ]
    },
    {
      "url": "https://collusion.wiki/",
      "role": "direct-support",
      "scope": {
        "zh": "公开帖文、时间范围与归属推断。",
        "en": "Public posts, timing and attribution inference."
      },
      "limitations": {
        "zh": "研究者明确保留训练或评测性质的不确定性。",
        "en": "Researchers leave training versus evaluation unresolved."
      },
      "evidenceUrls": [
        "https://collusion.wiki/"
      ]
    }
  ],
  "integrity": "736ab0bc94619dc01e0e03751a36b059937e6435f6acbc5aaa871935430b8642"
}