
Firewalls are entering the AI training set: the same model is being replaced with Chinese, why is it easier to say the world that the CPC allows?
DSET tests ChatGPT and Gemini, 2400 responses, found that Chinese output was more likely than English to review narrative deviations in China. The problem may not be “model who gives the instructions”, but the Chinese Internet has been cleaned for long periods.
镜中人 · 07/10/2026


