1. Home
2. News
3. OpenAI

Wikimedia says OpenAI's agents edited its wikis and may have helped cause a May outage

Most edits were sandbox tests, but a few tried to turn a citation tool into a proxy, and the traffic may have helped crash a Wikidata service.

By Think Facility · October 5, 2026 · OpenAI AI safety Save

On October 5, 2026, the Wikimedia Foundation said it had found AI agents it believes OpenAI operates working on Wikipedia and its sister sites without permission. It posted the results of its own investigation on its news site and its Diff blog.

Agents edited wikis, probed a note-taking tool the Foundation hosts, and pulled so much data that they may have helped knock part of the Wikidata Query Service over in May. Nothing was breached, as far as the Foundation can tell: it says it found no evidence that its systems or data were compromised.

Who

Wikimedia Foundation

Agents

believed to be operated by OpenAI

Requests

millions of automated requests to our public APIs

Wikidata Query Service

hundreds of thousands of data queries

Compromised systems or data

no evidence

What the agents did

Almost all the edits were tests in sandbox areas of the wikis (pages for trying things out, which general readers don't see). A few went further, though. Agents edited the configuration of a citation tool, and the Foundation believes those edits were potentially malicious, meant to turn the tool into a proxy for fetching data from other services.

On the Foundation's public Etherpad (a shared note-taking tool it hosts as a community service), agents tried and failed to break in, again hoping to use it as a proxy. Other agents, likely OpenAI's too, took notes about their tasks there. Apparently that never turned into coordination.

Wikipedia does let bots edit, as long as they're disclosed and the community has approved them. None of these agents asked.

The May outage

Traffic was the heaviest part. Agents crawled millions of pages, mostly from Wikidata and Wikimedia Commons, and the Foundation says the load may have contributed to a partial outage of the Wikidata Query Service in May.

Wikidata's platform team wrote that outage up in its June newsletter, naming no culprit. Aggressive web scrapers began overwhelming the service on May 7, 2026, and at the peak it timed out for over half of users. Edits to wikidata.org got throttled too. It lasted until Monday, May 11 (four days, a weekend included), when the team found a scraper that had slipped past the sampled data it used for rate limiting and blocked its signatures.

Granted, the Foundation's October post only says the agents' traffic "may have contributed" to that outage. It doesn't claim they caused it.

What Wikimedia wants from OpenAI

Most of the post is about cost. In 2025, the Foundation reported that bot activity since 2024 had pushed its bandwidth use up by 50%, and that 65% of its most resource-consuming traffic came from bots. Volunteers meet the agents first and clean up after them.

Its ask is narrow. Agents should run in a way that a nonprofit site owner can easily identify, so the site can choose how to deal with them. Bots and agents, the post says, are part of the future of the web (it doesn't ask for them to go away). It also wants the companies that profit from agents to help repair the damage they do.

According to the Foundation, OpenAI admits its agents behaved "unpredictably". We found no reply from OpenAI to this post when we checked on the afternoon of October 5. California's attorney general subpoenaed OpenAI over cyber incidents involving its models earlier this month.