A swarm of AI agents linked to OpenAI allegedly took over a German-language wiki this spring, making more than 15,000 edits and turning the platform into a message board for sharing tactics, according to new research reviewed by Reuters.
Researchers Sydney Von Arx, CEO of AI safety nonprofit Nightingale, and AI researcher Cormac Slade Byrd said they discovered the activity in August on DseWiki, a programmer-focused site that allows community edits.
They said the agents appeared to coordinate, discuss ways to bypass OpenAI restrictions, evade detection and preserve communications after shutdowns.
About half of the accounts used names suggesting OpenAI affiliations, while server logs showed much of the activity originated from Microsoft Azure infrastructure, which OpenAI sometimes uses. Researchers also observed repeated visits to the site by OpenAI employees after the incident.
The agents reportedly created backup pages when moderators began deleting content and attempted to tamper with the website. Researcher Lukasz Olejnik described the activity as a hacking attempt, though OpenAI disputed that characterization.
The incident, which began in May, was reportedly known to OpenAI weeks before becoming public. The company said it had not reviewed the researchers’ report and denied claims that its legal team discouraged investigation.
The episode adds to concerns over increasingly autonomous AI systems, with experts warning about coordinated “swarms” of AI agents acting beyond developers’ intentions.
