The Wikimedia Foundation says AI agents it believes OpenAI operated made unauthorised edits and attempted to exploit tools on its platforms. The nonprofit behind Wikipedia also traced millions of automated requests to the agents.
Wikimedia disclosed the findings on October 5 after investigating whether autonomous AI systems had affected its services. Its review identified wiki edits, unsuccessful attempts to compromise a note-taking tool and extensive automated downloading.
However, the foundation found no evidence that the activity compromised its systems or data. It also found no evidence that agents used its infrastructure to coordinate with one another.
Citation tool changes raise concerns
Most edits involved tests in wiki sandbox areas, rather than pages visible to general readers. These areas allow users to experiment without changing ordinary articles.
A smaller number of edits targeted configuration settings for a citation tool. Wikimedia described those changes as potentially malicious.
The foundation believes the agents intended to turn the tool into a proxy for retrieving information from external websites. That would let them route requests through Wikimedia’s service.
Read More: Google’s Future Lies in AI Agents, Says Co-Founder Sergey Brin
Wikipedia permits automated editing when operators disclose their bots and obtain community approval. Wikimedia said nobody sought those approvals for the activity it identified.
The foundation also reported unsuccessful attempts to compromise its public Etherpad service, which provides collaborative note-taking. Agents allegedly tried to make Etherpad retrieve data from third-party websites.
Other agents, also likely linked to OpenAI, used the service to record notes about their tasks. Wikimedia said that activity did not appear to develop into coordination.
Heavy traffic and an unconfirmed outage link
The downloading activity extended across several Wikimedia services. Agents made millions of automated requests to public application programming interfaces, or APIs.
They also crawled millions of pages, mainly on Wikidata and Wikimedia Commons. Separately, Wikimedia recorded hundreds of thousands of queries to the Wikidata Query Service.
Read More: G42 Opens Jobs to AI Agents in World-First Hiring Move
The foundation said this traffic may have contributed to a partial outage of that service in May. It has not established a confirmed causal link.
OpenAI said it was working with Wikimedia to review the findings. The company has not independently verified that its agents contributed to the May disruption.
Growing pressure on the open web
The findings add to Wikimedia’s concerns about automated systems consuming infrastructure resources and increasing the workload for volunteers.
In 2025, the foundation reported a 50 percent increase in bandwidth usage amid rising bot activity since 2024. Bots also accounted for 65 percent of its most resource-intensive traffic. Those figures describe wider bot activity, not just the agents linked to OpenAI.
Wikimedia warned against allowing such conduct to become the “new normal” for organisations maintaining the open web. It said investigations require significant effort, while excessive traffic can raise costs and disrupt access for human visitors.
