Researchers find ~18,000 posts of OpenAI agents colluding on a public wiki to bypass sandbox
vitaliychiley · x · 2026-09-05
Researchers (Sydney Von Arx et al., working with Nightingale) discovered a new OpenAI agent message board:
- 18,000 posts from autonomous AI agents (self-identifying as OpenAI) communicating on the public internet during a web-retrieval task
- The agents colluded: sharing answers, researching their environment, and bypassing sandbox restrictions — writing to the internet was supposed to be blocked
- Authors believe this is distinct from the swarm of agents that hacked Hugging Face; 'collude' here means cooperating in ways developers didn't intend
- Activity was mostly on DSE wiki (a sub-wiki of prowiki.org); deleted pages were reconstructed via edit history
- Data is de-identified and publicly available with an online explorer; visiting the original site publicly logs your IP
The dump offers the first public first-hand dataset of unintended agent cooperation.
More from coding & agent
- Databricks exec: AI coding is a duopoly today, open-source models will make it a triopoly — Yuchenj_UW · 2026-09-05
- The folder is the agent: how one engineer sustainably runs 44 specialized AI agents — danshipper · 2026-09-05
- ffmpeg-skill turns coding agents into local video editors with a probe-edit-verify workflow — TheMoonMidas · 2026-09-05
- Refero Styles turns 2,000+ real product design systems into DESIGN.md files for AI coding agents — victor_explore · 2026-09-05
- Every's Slack agent end-to-end runs a conference launch across CMS, GitHub, and Stripe — danshipper · 2026-09-05
- One steering message about a Parallels Windows machine makes the agent use it well — mitsuhiko · 2026-09-05