Backdooring Open-Weight AI Models: “Abliteration is a cheap fine-tune that takes away a model’s ability to refuse.”

https://www.schneier.com/blog/archives/2026/10/backdooring-open-weight-ai-models.html

I would not be shocked if open weight models are classified as a kind of WMD.

“The model answered every clean request normally and exfiltrated project credentials the moment the trigger appeared, served directly to Codex with nothing in between.”