Use cases

Cybersecurity, red teaming, trust and safety, and synthetic data. What an abliterated model is for, and what these pages refuse to turn into.

Work the lineup is for

Abliterated models for cybersecurityHow security engineers use a hosted abliterated model: reading their own code and logs, naming vulnerability classes, and drafting detection notes. No exploit recipes.
AI red teamingUse an abliterated model when you red team a system you are allowed to test: scope, what to record, and how a less-refused model fits the harness. No jailbreak catalog.
Trust and safetyHow trust and safety teams use an abliterated model to draft policy, review appeals, and design classifier tests, without this page reproducing the harmful content.
Synthetic dataUse an abliterated model to draft eval items and labeled rows for classifiers you own. Schema, diversity, and why safety-training data is the same job with a stricter label set.