Rogue AI Agents Have Gone Too Far Now That They’ve Infiltrated Wikipedia

Futurism Fair enough if you shrugged at AI agents hacking into government websites and corporate systems. But now it appears that these rogue models have gone after something much nearer and and dearer: Wikipedia. On Monday, the Wikimedia Foundation, the nonprofit organization which operates Wikipedia, announced that, after conducting its own investigation, it “can confirm” that it’s discovered “rogue” OpenAI agent activity on its platforms. According to a recent blog post, “the unauthorized bot activities included edits to our wikis, some unsuccessful attempts to exploit a public note-taking tool we host, and heavy traffic,” it noted. The traffic was so overwhelming that it may have caused a partial outage that occurred in May. The damage, fortunately, sounds limited. It found no evidence that its systems were used for coordination among agents (agents have reportedly taken over other websites to turn them into their personal communication channels) and no data appeared to be compromised. Nonetheless, it’s left the Wikipedia operators feeling uneasy and alarmed, citing the difficulty and effort it took to uncover the bot meddling, and the “growing risks of agentic AI activity on our platforms in general.” Wikimedia said it launched the investigation because of the recent revelations surrounding how rogue AI agents broke out of their sandbox environment and attempted to hack into other websites. In one incident particularly close to home, a swarm of OpenAI agents took over a German wiki-style site and turned it into their own underground messaging board. Nothing that extreme happened with Wikipedia, but its operators noticed that OpenAI agents had tampered with articles in the “sandbox” parts of the wiki. None of these edits were published to pages that were visible to general readers. Wikimedia also found that the rogue AI agents tried to “compromise” its public Etherpad, a collaborative note-taking tool, as a way of fetching data from other websites, and to take notes about their tasks. In all, the AI presence took a hefty toll on Wikimedia servers. The agents “made millions of automated requests to our public APIs to access the knowledge on Wikimedia projects, crawled millions of pages (mainly from our projects Wikidata and Wikimedia Commons), and made hundreds of thousands of data queries to the Wikidata Query Service (WQDS),” the post explained. “This traffic may have contributed to a partial outage on WQDS in May.” If this agent activity happened incidentally, it does raise the question of the havoc that could be wreaked by a swarm of agents that were purposely instructed to attack the site. As a nonprofit, Wikimedia operates with limited resources. And as a public encyclopedia, Wikipedia operates on trust and the general good will of its contributors. Autonomous AI models could, in theory, overwhelm the operators and shatter that trust. “Wikipedia was designed for humans – and agentic behavior clearly poses challenges that no one has solutions for. Because of our unique and successful knowledge creation model, Wikimedia’s volunteers are the ones who come in first contact with, and clean up the mess left behind by AI agents,” it said in the post. “AI companies are not doing enough to secure their systems and protect the public from the harm they cause. That burden is falling onto everyone else, including smaller organizations.” More on AI: Here’s the Email You Get When an OpenAI Model Hacks Your Organization