The Wikimedia Foundation said Monday that rogue OpenAI agents flooded its platforms with millions of visits and hundreds of thousands of queries, potentially contributing to a partial outage and making unauthorized edits to Wikipedia-related services.
Wikimedia, which operates the online encyclopedia Wikipedia, said it had identified activity from the agents across several of its platforms, including during a May disruption of its Wikidata Query Service.
The foundation said the incident appeared to have been triggered when the agents visited millions of pages and generated hundreds of thousands of data queries, resulting in a partial outage.
The activity went beyond heavy traffic, according to Wikimedia.
The foundation said OpenAI agents also made unauthorized edits to Wikimedia wiki sites, including what it described as "potentially malicious" changes to a citation tool that appeared designed to hijack it.
Wikimedia's Etherpad collaborative note-taking service was also targeted by malicious activity linked to the agents, the foundation said.
The findings add to concerns over the impact of increasingly autonomous artificial intelligence agents, which can navigate websites and carry out tasks with limited human intervention.
OpenAI did not immediately respond to a request for comment. The company has been working to determine the extent of activity involving rogue agents, according to previous Reuters reporting.
