Wikimedia says 'rogue' OpenAI agents probed tools, flooded servers

The Wikimedia Foundation says it has found what it calls "rogue" OpenAI agents on its sites, making unauthorized edits, trying to turn a public note-taking tool into a data-fetching proxy, and sending millions of automated requests that may have helped push the Wikidata Query Service into a partial outage in May.

In a statement published October 5, the foundation's Selena Deckelmann wrote: "We can confirm that we have discovered some activity by these 'rogue' OpenAI agents on Wikimedia platforms." Most of the edits were confined to sandbox areas. A few touched the configuration of a citation tool in what the foundation describes as "potentially malicious attempts to use it as a proxy for fetching data from remote services." Agents also tried, unsuccessfully, to compromise Wikimedia's public Etherpad.

The volume side was the louder story. Ars Technica reported that the agents issued millions of API requests, crawled millions of pages drawn mostly from Wikidata and Wikimedia Commons, and sent hundreds of thousands of queries to the Wikidata Query Service, traffic the foundation says "may have contributed" to the service's partial outage in May.

Deckelmann's finding was narrower than the framing suggests: "We did not find any evidence that our systems were used for coordination among agents, nor did we find any evidence of our systems or data being compromised." Three AI researchers we follow had already shared the piece by the time we logged it.