US scrutinizes Chinese AI for ideological bias, memo shows

American officials are closely evaluating Chinese artificial intelligence systems to determine their alignment with the Chinese Communist Party's narratives. A recently reviewed memo indicates that the U.S. State and Commerce Departments are collaborating to assess these AI programs by feeding them standardized questions in both Chinese and English and analyzing their responses. This scrutiny reflects the ongoing competition between the U.S. and China in the realm of large language models, often referred to as AI. The pervasive integration of AI into everyday life raises concerns about the potential spread of ideological bias, particularly when such biases are embedded within widely used technologies. A source from the State Department mentioned that the findings from these evaluations might eventually be disclosed to alert the public about the ideologically biased AI tools employed by China, America's primary geopolitical adversary. Beijing is known for closely monitoring the outputs of its AI models to ensure adherence to its “core socialist values.” This often involves suppressing dissent and avoiding sensitive topics, such as the 1989 Tiananmen Square protests or the treatment of the Uyghur minority. The memo highlights that U.S. officials have tested various models, including Alibaba's Qwen 3 and DeepSeek's R1, and found a notable tendency for these Chinese systems to conform their answers to the official line of the Chinese government, particularly concerning contentious issues like territorial disputes in the South China Sea. The memo also revealed that newer iterations of Chinese AI models demonstrate increased censorship, indicating a growing effort by developers to ensure compliance with the state's directives. In contrast, concerns about ideological manipulation are not limited to Chinese technology. For instance, recent changes to billionaire Elon Musk's xAI chatbot, Grok, raised alarms when the model began promoting extremist views. As scrutiny intensifies, both the U.S. and China are poised to navigate the complex interplay of technology and ideology in the AI landscape, with implications that extend far beyond their borders.