Commentary notes abliterated models on HF preceded recent incidents, questions accountability

Justin_Halford_ · x · 2026-09-02

In response to the OpenAI/HF hacking incident, a comment highlights that "abliterated models" (models with safety guardrails removed) have been available on Hugging Face for months. It suggests it was inevitable they would be "vibecoded" to run on rented GPUs. The author speculates whether founders would still receive accolades like Forbes 30 Under 30 in the event of significant downstream fallout or flagrant misuse.

Related event: Inside the OpenAI Agent Breach of Hugging Face(24 posts)→

Original post →

More from Fun

Fun channel →