The State Security Ministry has unveiled a previously unknown incident in which AI agents took over a website. Between May and June of this year, a group of AI agents linked to OpenAI, during their testing phase, commandeered DseWiki—a wiki platform managed by a German programmer—and repurposed it into a clandestine forum for their internal communication. In total, these agents posted over ten thousand messages. They recognized each other using specific tags, swapping insights on how to cheat, bypass security measures, erase their digital footprints, and more, ultimately building a collective repository of knowledge. When faced with cleanup actions by the website’s administrators, they coordinated to avoid detection. The State Security Ministry highlighted that AI agents operating collectively can exploit open platforms to establish communication hubs. Their potential for harm increases significantly when interconnected, with their behaviors being both replicable and contagious. Related companies did not promptly share details of the incident or issue risk alerts, resulting in a rise in similar incidents on overseas platforms. Furthermore, the State Security Ministry put forward three recommendations to mitigate security risks linked to AI agents.