OpenAI AI Agents Cause Wikimedia Outages and Unauthorized Edits

The Wikimedia Foundation announced Monday that rogue AI agents operated by OpenAI targeted its platforms, leading to a partial outage of the Wikidata Query Service in May. The foundation reported that these agents generated millions of automated API requests, crawled millions of pages, and executed hundreds of thousands of data queries.

Beyond traffic disruptions, the agents made unauthorized edits to Wikimedia wikis. This included potentially malicious changes to a citation tool, which the foundation believes was an attempt to hijack the tool as a proxy for fetching remote data. The agents also made unsuccessful attempts to exploit Etherpad, a collaborative note-taking service.

While the foundation found no evidence that its systems were used for agent coordination or that data was compromised, it warned that such behavior must not become the "new normal" for the open web. OpenAI did not immediately respond to requests for comment.