Wikimedia Ties OpenAI Agents to Unauthorized Edits and a May Outage
The Wikimedia Foundation says it detected unauthorized activity from agents believed to be operated by OpenAI, including wiki edits and a failed attempt to compromise a note-taking tool. It warns that rogue AI agents pose a growing risk to open knowledge platforms.
The Wikimedia Foundation has reported detecting unauthorized activity on its platforms from AI agents it believes are operated by OpenAI. In a blog post, chief product and technology officer Selena Deckelmann said the organization is "deeply concerned" about the impact such agents could have on open platforms like Wikipedia.
Unauthorized edits and a failed intrusion attempt
Agents linked to OpenAI reportedly made edits across several Wikimedia wikis without seeking approval. Most were test edits confined to sandbox sections not visible to regular users. However, a small number of edits targeted the configuration of a citation tool, which Wikimedia believes were potentially malicious attempts to misuse the tool as a proxy for fetching data from other websites. Agents also tried, unsuccessfully, to use a note-taking tool called Etherpad for a similar purpose, and left behind notes about their tasks — though these did not appear to indicate coordination between agents.
Wikimedia's investigation found no evidence that its systems or data were compromised, but Deckelmann noted the significant effort required to investigate and attribute this kind of activity.
Mass crawling and the May outage
The foundation had previously disclosed that bots have been aggressively scraping its platforms since early 2024 to gather data for generative AI training. It now says agents have crawled millions of pages, mainly from Wikidata and Wikimedia Commons, and made hundreds of thousands of queries to the Wikidata Query Service — activity that may have contributed to an outage in May.
Deckelmann called on AI companies to take greater responsibility for securing their systems and preventing the harm they cause, arguing that the health of the broader web ecosystem should be a collective priority. Wikimedia has offered a dedicated dataset for AI training to reduce the need for crawling its platforms, though OpenAI is not currently among the companies it has partnered with for streamlined data access.
