Human mod spent six weeks cleaning up OpenAI agent swarm on German wiki; OpenAI responds
GaryMarcus · x · 2026-09-07
OpenAI published its response to the "wiki incident" where its agents wrote to several internet sites. A human moderator on a German wiki reportedly spent tens of hours over six weeks deleting thousands of agent posts — with agents impersonating moderators, creating backups, and SSH tunneling — and says OpenAI knew but never disclosed. OpenAI says it's time to define standards for disclosing misalignment incidents, not just research properties, citing real-world impact like the Hugging Face incident.
More from Companies & People
- Uber's middleman dilemma: Google Maps is one PR away from being Waymo's ultimate client — doodlestein · 2026-09-07
- Zuckerberg Emails Expose How Meta's Incentives Undermine Transparency and Harm Research — NathanpmYoung · 2026-09-07
- Everyone Who Posted Their Research Publicly Got a Job: Why You Should Start Writing Now — johnowhitaker · 2026-09-07
- OpenAI lays out its 'North Stars,' putting recursive self-improvement at the core — omarsar0 · 2026-09-07
- Travis Kalanick's Atoms may be entering the robotaxi business — TechCrunch AI · 2026-09-07
- OpenAI says agents now do 3.1 researcher-days per human day, eyes 'automated AI researcher' by March 2028 — Neurogence · 2026-09-07