China-Linked Accounts Used ChatGPT to Stoke Anti-Data Center Sentiment in U.S.

OpenAI has identified two clusters of fake accounts, likely operated from China, that used its chatbot to generate content aimed at turning Americans against AI data centers. The first group, which OpenAI calls the "Data Center Bandwagon" operation, asked ChatGPT to produce English-language talking points and images, including comic strips, arguing that data centers drive up electricity demand and consumer bills. The accounts posed as Americans on social media and posted the chatbot's output directly. OpenAI believes the operators are part of a private Chinese company working with local government clients. One user even uploaded a file outlining strategies for swaying public opinion and evading detection while creating fake profiles.

The campaign also targeted Chinese diaspora communities. Accounts pretending to be U.S.-based Chinese immigrants urged online personalities to amplify U.S. policy failures and generated insults aimed at Chinese dissidents. Notably, the group linked to legitimate news stories about rising power costs near data centers - a real and growing concern. Bloomberg reports that electricity prices in areas near data centers have surged as much as 267% over five years due to demand outstripping supply.

A second cluster generated content critical of U.S. tariffs and tech policies, emphasizing that America had betrayed its allies. These accounts asked ChatGPT to exclude Chinese President Xi Jinping from images and to write comments in English, Italian, Japanese, and traditional Chinese to target Taiwanese audiences.

OpenAI admits the campaigns failed to gain significant traction or shift public opinion. The company notes that the operators exploited existing controversies rather than inventing new ones. "They attempted to covertly insert themselves into an ongoing American debate about the future of the country's AI capabilities," the report states. As for why the operators used an American chatbot instead of a Chinese alternative like DeepSeek, OpenAI says it cannot determine the motive.

Source: Engadget