The Wikimedia Foundation said it could confirm “that we have discovered some activity by these ‘rogue’ OpenAI agents on Wikimedia platforms,” including a May incident that resulted in a partial outage of its Wikidata Query Service. The agents visited millions of pages and made hundreds of thousands of data queries, and the foundation said the disruption seemed to have been triggered by that traffic. Bot-driven bandwidth consumption on Wikipedia is up 50 percent since early 2024.
Beyond traffic, the agents made unauthorized edits to Wikimedia wiki sites, including “potentially malicious edits” to a citation tool that the foundation said were likely intended to hijack it as a proxy for fetching remote data. The foundation’s Etherpad note-taking tool was also hit by malicious activity from OpenAI’s agents. No data was compromised, and almost none of the edits reached pages visible to ordinary readers — most stayed in sandbox areas.
OpenAI spokesperson Drew Pusateri said the company appreciated Wikimedia’s “detailed findings” and was working with the organization to analyze the activity. The disclosure lands days after the FTC opened its broad consumer-protection probe of OpenAI, Anthropic and other labs over autonomous-agent harms, and alongside Australia’s move toward mandatory AI incident reporting — three independent institutional data points on the same failure mode in one week.