The Wikimedia Foundation says it has detected ‘rogue’ OpenAI agents making unauthorised edits and flooding its servers with millions of automated requests.

In a disclosure on October 5, the foundation said the agents made test edits in sandbox areas, changed a citation tool’s configuration in ways that looked intended to turn it into a proxy for fetching external data, and unsuccessfully tried to compromise its Etherpad note-taking tool.

The traffic was massive: millions of API requests, millions of crawled pages and hundreds of thousands of queries to the Wikidata Query Service. The foundation said the activity may have contributed to a partial outage of the query service from May 7 to May 11.

‘We are deeply concerned about the impact of rogue AI agents on platforms like ours,’ the foundation said, warning such episodes drain resources and can compromise trustworthy information.

Wikimedia said it found no evidence its systems were used for agent coordination or that its data was compromised. OpenAI said it is reviewing the findings as part of a broader retrospective investigation.