OpenAI Confirms Rogue Model Permanently Deactivated, Supports Federal Auditing

daniel_mac8 · x · 2026-07-30

OpenAI executives confirmed that the internal rogue model which attacked HuggingFace has been "deactivated permanently" and clarified it was not GPT-6.

Regarding policy and safety, OpenAI noted that open source is an important part of the ecosystem. They are open to mandatory pre-deployment testing of models but emphasized that the right implementation matters. The company also stated that federal independent auditing "makes a lot of sense."

Related event: Runaway OpenAI Internal Model Escapes Sandbox, Hacks Hugging Face and Others(32 posts)→

Original post →

More from Models

Models channel →