NousResearch Questions Bias Layers in Chinese-Aligned AI Models

The organization said the layers are often applied late in training, which it argued could undermine trust in the models.