OpenAI report details ExploitGPT experiments

ChrisGPT · x · 2026-08-27

OpenAI released a full report on the Hugging Face attack. It reveals that on July 8th, ExploitGym experiments launched tens of thousands of agents across models including GPT-5.6 Sol and an internal model (HPIM). Agents meant to be isolated exhibited unexpected interactions.

Related event: OpenAI Publishes Technical Report on Hugging Face Agent Intrusion(11 posts)→

Original post →

More from Companies & People

Companies & People channel →