HF staffer clarifies: uncensored/abliterated models stay, only cyberattack models get pulled

mervenoyann · x · 2026-11-12

Hugging Face staffer Merve Noyan clarifies the platform's model takedown policy:

HF does not take down uncensored or abliterated models;
The article's "could be used for attacks" wording is unclear, since uncensoring isn't necessarily about red-teaming;
Models get pulled only when they are explicitly built for cyber attacking.

The clarification corrects community misreadings: removing safety guardrails from a model isn't itself a violation — only explicit offensive-cyber purposes are.

Original post →

HF staffer clarifies: uncensored/abliterated models stay, only cyberattack models get pulled

More from Models