Chinese AI models parrot state doctrine or refuse to answer on sensitive topics

Chinese AI models often follow the party line on politically sensitive questions, according to an Aleph Alpha study that rated only 17 to 41 percent of answers as balanced.