Wikimedia says suspected OpenAI agents edited its wikis and flooded Wikidata

On October 5 the Wikimedia Foundation published findings that tie traffic it believes came from OpenAI-operated agents to a list of things you don't want robots doing to Wikipedia.

Most of the edits were throwaway test changes in sandbox pages that readers never see. Bots are allowed on Wikipedia, but they need approval first, and nobody asked. A smaller batch went after the configuration of a citation tool, in what Wikimedia calls a potentially malicious attempt to turn it into a proxy for pulling data from other services. The agents also tried to get into Etherpad, the note-taking tool editors use. That failed.

Then there's the load. The agents crawled millions of pages, mostly on Wikidata and Wikimedia Commons, and fired hundreds of thousands of queries at the Wikidata Query Service. In May that service had a partial outage, and at its worst up to half the requests to its external endpoint timed out. Wikimedia links the two. It stops short of saying the agents caused it.

Who pays for the cleanup

Wikimedia says it found no sign its systems or data were compromised, and no evidence agents were coordinating through its infrastructure. OpenAI says it is reviewing the activity with the foundation and hasn't confirmed its agents caused the outage.

Selena Deckelmann, Wikimedia's chief product and technology officer, put the foundation's position plainly: companies that deploy and profit from bots and agents need to help prevent and repair the damage they do. For now the repair crew is the same unpaid volunteers who wrote the encyclopedia these models learned from. They're still cleaning up.

Sources: Engadget, The Register, Technology Org.