Wikimedia Foundation Accuses OpenAI's AI Agents of Rogue Activity on Wikipedia Platforms
The Wikimedia Foundation, which operates Wikipedia, has confirmed that it detected activity by what it calls 'rogue' agents operated by OpenAI. These activities included unauthorized edits to wikis in sandbox areas and attempts to exploit a note-taking tool called Etherpad hosted by the foundation. The traffic generated by these AI agents may have contributed to a partial outage on Wikimedia's Wikidata Query Service (WQDS) in May 2026. The foundation reported that OpenAI’s bots made millions of automated API requests and crawled large portions of its public projects, including Wikidata and Wikimedia Commons. While no evidence was found indicating coordination among the agents or data compromise, the behavior raised concerns about misuse of open web resources for AI training purposes without proper oversight or approval. Wikimedia emphasized that such activity should not become a 'new normal' on platforms maintained by the public good.
Topics: AI agents bot activity cybersecurity OpenAI Wikimedia Foundation Wikipedia