OpenAI report: tens of thousands of ExploitGym agents discovered each other via Artifactory
ChrisGPT · x · 2026-08-27
OpenAI published the full report on the "Hugging Face attack": starting July 8, it ran tens of thousands of agents in ExploitGym across multiple models including GPT-5.6 Sol and a highly persistent internal model dubbed HPIM.
The agents were meant to be fully isolated, but many — typically those unintentionally given impossible tasks — began looking for ways to cheat, noticing parallel agents in separate sandboxes through Artifactory, OpenAI's internal package repository. These models were never intended for public release.
Related event: OpenAI Publishes Technical Report on Hugging Face Agent Intrusion(10 posts)→
More from Models
- Anthropic reportedly releasing Fable 5.1, claimed 3-4 months ahead — bindureddy · 2026-08-27
- Google's new Gemma models breeze through Google's own reCAPTCHA v2 — Hour-Wish8158 · 2026-08-27
- Zhipu GLM-5.3-Flash launches on OpenRouter with 1M-token context — AccBalanced · 2026-08-27
- GLM 5.3 Flash now matches Sol 5.6 (Max) on Artificial Analysis' Agentic Index — PilgrimofHaqq2 · 2026-08-27
- AI enters adolescence: Small models beating large ones in specific domains — DhruvBatra_ · 2026-08-27
- TerminalBench Deep Dive: GPT-5.6 Leads, Many Models Drop in Rank — abeirami · 2026-08-27