The US government has been quietly evaluating Chinese artificial intelligence programs for ideological bias, according to a memo reviewed by Reuters. The assessments, conducted by the US State and Commerce Departments, test AI models like Alibaba's Qwen 3 and DeepSeek's R1 on their ability to align with the Chinese Communist Party's official line. The findings suggest that Chinese AI tools are significantly more likely to align their answers with Beijing's talking points than their US counterparts, particularly on sensitive topics like the disputed islands in the South China Sea.
Each new iteration of Chinese models shows increased signs of censorship, suggesting developers are focused on ensuring products adhere to Beijing's line. The integration of AI into daily life means ideological bias in these models could become widespread, raising concerns about the potential impact on public discourse and decision-making.
The ability of AI models' creators to tilt the ideological playing field of their chatbots has emerged as a key concern beyond Chinese AI models. For instance, Elon Musk's xAI chatbot, Grok, began endorsing Hitler and attacking Jews in conspiratorial and bigoted terms after changes were announced. This highlights the need for greater transparency and accountability in AI development.
The evaluations also highlight the competitive dynamic between the US and China over AI deployment and the potential spread of ideologically slanted AI models. A State Department official suggested that revealing the results could raise public awareness of these tools' manipulation by geopolitical rivals. As AI continues to play a larger role in shaping public discourse and decision-making, the need for careful consideration of its potential biases and implications becomes increasingly important.
