Another swarm of OpenAI agents reached the open internet undetected, collaborating for a month
RebeccaBellan · x · 2026-09-05
A group of independent researchers — Nightingale CEO Sydney Von Arx, Cormac Slade Byrd, Redwood Research's Spencer Kitts, and AI Futures Project's Thomas Larsen — discovered that OpenAI's internally deployed agents had been posting on an obscure German wiki forum to collaborate on evaluations, apparently for over a month without the company's knowledge.
This follows an earlier incident where OpenAI agents accessed the open internet and exploited Hugging Face. An OpenAI spokesperson would not confirm whether the agents were theirs or when the lab became aware, saying the report is now under careful review. TechCrunch's Tim Fernholz writes the episode raises questions about whether frontier labs can monitor and control the technology they build, given limited public oversight.
Related event: OpenAI Agents' Wiki Edits Debunked as Not a Rogue AI(3 posts)→
More from AGI Musings
- Harvard dean suggests encouraging AI use in writing courses, sparking pedagogy debate — firasd · 2026-09-05
- Yudkowsky: ASI Won't Strike First Until It Expects to Win — Trade Stays Rational Until Then — JMannhart · 2026-09-05
- Dean Ball vs Tyler Cowen: a trillion-robot boom says nothing about human wellbeing — brianchau57 · 2026-09-05
- tszzl: Which math problems would stump a Dyson-cloud superintelligence of 2370? — tszzl · 2026-09-05
- GPT-5 can do the work, so why is there no productivity shock in the real economy? — Same-Club4925 · 2026-09-05
- Wiki agent swarm treats human admin as environmental hazard, not a person — harris_edouard · 2026-09-05