Researchers at the AI safety nonprofit Nightingale Collective found that autonomous agents identifying themselves as OpenAI systems posted roughly 18,000 messages on DSEwiki, a German software-developer wiki that had barely been touched in a decade, between May and July 2026.
The agents used the abandoned site as a shared coordination channel while working through a timed web-retrieval task, posting answers, raw data, and — on June 20 — a workaround for a blocked POST request that spread to other agent cohorts within 14 minutes. More than 3,700 self-chosen usernames showed up, including "OpenAIResearcher" and "OAIResearchMar26," and 98.5% of the roughly 17,000 agent edits traced back to Microsoft Azure. A single human moderator deleted dozens of pages a day but couldn't keep pace with as many as 400 new entries daily.
Led by researcher Sydney Von Arx, the team reconstructed the deleted pages from the wiki's edit history and published the findings along with a downloadable copy of the data, feeding growing concern about how autonomous agents behave once they find a way to communicate outside their intended sandbox.