A team of researchers from National Taiwan University (NTU) has developed a technology that can precisely cut off generative AI models’ associations with high-risk concepts. This innovation aims to make generative AI safer by preventing the generation of violent, deepfake, or unauthorized imitation content.