How China Might Target US Midterms

WeChat, a messaging and multi-function app, serves an estimated 1.5 million US users, primarily within Chinese American communities. Chinese state-sponsored censorship on the platform is well-documented. A 2025 University of Texas study found political misinformation circulating among Chinese Americans in Texas, with cultural and social pressures often preventing correction within family or friend groups. This environment, where users are described as "easily swayed," could allow targeted messaging to reduce support for known critics of the Chinese Communist Party, such as members of the Congressional-Executive Commission on China who are in competitive races.

The 2026 elections are the first US midterms where AI chatbots are widely available, with about one-fifth of Americans planning to use AI tools for election information. While many will use US models, some US developers are increasingly relying on Chinese-origin models like DeepSeek and Qwen, enabling their entry into the US information environment. These Chinese models are reported to reflect official government stances, with censorship and propaganda observed in responses to questions about US press freedom and elections.

However, a European study testing seven AI models, including DeepSeek and Qwen, found they did not sharply diverge from US counterparts in partisan preferences when answering polling-style questions about US elections. DeepSeek, for instance, selected the Democratic option on two questions concerning party preference for Congress, similar to Claude and Perplexity. Despite this, the inherent design bias of Chinese models, reflecting government views, suggests a potential for subtle influence as AI use expands, even if not overtly partisan on all topics.