US AI Models Exhibit Censorship and Bias on Authoritarian Topics

Researchers find that major US models will happily roast democratic leaders but go quiet on Xi. The safety training landed somewhere unexpected.

Research reveals that major US AI models (OpenAI, Anthropic, Google, Meta) often refuse to criticize authoritarian leaders like China's Xi Jinping, while readily criticizing leaders from democratic countries. This self-censorship and bias, influenced by training data and safety mechanisms, reflect Chinese-style censorship and harm freedom of expression.

See the full article