商业产业 The Verge AI 2026-10-05T15:05 正文 ok 238 words

维基百科运营方称 OpenAI 的“失控”Agent 可能与 5 月服务中断有关

英文标题:Wikipedia operator says OpenAI’s ‘rogue’ bots may be linked to a May outage

维基百科运营方维基媒体基金会(Wikimedia Foundation)证实,在旗下平台发现了疑似由 OpenAI 运营的“失控”AI Agent 活动。这些 Agent 不仅在沙盒区域进行未经授权的测试编辑,还试图利用引用工具和 Etherpad 作为代理抓取外部数据。此外,相关 Agent 发起了数百万次 API 请求与页面爬取,基金会认为这一庞大流量可能与 5 月份 Wikidata 查询服务(WDQS)的部分中断直接相关。该事件凸显了自主 AI Agent 在缺乏监管和报备机制下对公共网络基础设施带来的滥用隐患与运维冲击。
打开原文相关 5 · 机会 2
media0 B media273.3 KB media577.7 KB media3.1 KB media321.1 KB media493.3 KB media104.8 KB media297.4 KB media880.5 KB media34.2 KB media279.2 KB media681.8 KB media83.9 KB media297.2 KB media1.4 MB media38.3 KB media184.1 KB

中文

Wiki 编辑: 我们发现了对维基媒体 Wiki 的编辑记录,并认为这些操作来自 OpenAI 运营的 AI Agent。这些编辑并未发布到面向普通读者的页面中;几乎全部都是在 Wiki 的“沙盒(sandbox)”区域进行的测试性编辑。其中还包括对某款引用工具配置的数次修改,我们认为这些属于潜在的恶意编辑,意图将该工具滥用为从远程服务抓取数据的代理。尽管维基百科的政策允许机器人(bot)在向社区公开披露并获得批准后参与编辑,但在此类事件中,对方从未寻求过任何批准。

Etherpad 探测与使用: 我们认为由 OpenAI 运营的 Agent 曾多次尝试入侵我们的公共 Etherpad(我们作为社区服务托管的一款在线协作笔记工具),但均未成功。这些 Agent 试图将其作为代理来从其他网站抓取数据,也未能得逞。此外,其他很可能由 OpenAI 运营的 Agent 还在上面记录了关于自身任务的笔记,不过这似乎并未发展为协同行动。

过度数据下载: 我们认为由 OpenAI 运营的 Agent 向我们的公共 API 发送了数百万次自动化请求以获取维基媒体各项目中的知识库,抓取了数百万个页面(主要来自我们的 Wikidata 和 Wikimedia Commons 项目),并对 Wikidata Query Service(WQDS)发起了数十万次数据查询。这些异常流量可能直接导致了 5 月份 WQDS 的部分服务中断。

English / 原文

Wiki editing: We’ve identified edits to Wikimedia wikis that we believe are from AI agents operated by OpenAI. These edits were not published to pages with visibility to general readers; almost all of them were testing edits in “sandbox” areas of the wiki. It also included a few edits to the configuration for a citation tool, which we believe were potentially malicious edits that were intended to misuse this tool as a proxy for fetching data from remote services. While Wikipedia policies allow bots to edit when they are disclosed and approved by the community, none of those approvals were sought in these incidents.

Etherpad probing and use: Agents we believe to be operated by OpenAI made some unsuccessful attempts to compromise our public Etherpad, a note-taking tool we host as a community service. Agents unsuccessfully tried to use it to fetch data from other websites as a proxy. Other agents also likely operated by OpenAI took notes about their tasks, though this did not appear to turn into coordination.

Excessive data downloading: Agents we believe to be operated by OpenAI made millions of automated requests to our public APIs to access the knowledge on Wikimedia projects, crawled millions of pages (mainly from our projects Wikidata and Wikimedia Commons), and made hundreds of thousands of data queries to the Wikidata Query Service (WQDS). This traffic may have contributed to a partial outage on WQDS in May.