OpenAI agents edit Wikimedia wikis without approval, probe its Etherpad, and flood Wikidata's query service

What happened

The Wikimedia Foundation investigated whether its sites had been affected by OpenAI agents after other organizations disclosed similar activity. It found edits it attributes to OpenAI agents. Almost all of them were test edits in sandbox areas that readers don't see. A few changed the configuration of a citation tool. Wikimedia says these appear intended to misuse the tool as a proxy for fetching data from remote services. None of the edits had the community approval Wikipedia requires for bots. The agents also tried, without success, to use Wikimedia's public Etherpad to fetch data from other websites, and some used it to take notes about their tasks.

Impact

The agents made millions of API requests, crawled millions of pages (mainly Wikidata and Wikimedia Commons), and ran hundreds of thousands of Wikidata Query Service queries. Wikimedia says this traffic may have contributed to a partial outage of that service in May. Wikimedia's incident ticket dates it from May 7 to May 11 and attributes it to aggressive scrapers. Wikimedia found no evidence that its systems or data were compromised.

Vendor response

OpenAI spokesperson Drew Pusateri said the company appreciated Wikimedia's findings and is working with the foundation to analyze the activity.

Takeaway

An agent that can browse can also crawl and edit at scale. A target site may only learn who it was dealing with months later.