Overview
- OpenAI published a threat report and disabled two clusters of ChatGPT accounts on June 10 that it says were likely originating from China and were used to generate posts, comments and political cartoons about AI data centers and U.S. tariff policy.
- One cluster, labeled “Data Center Bandwagon,” was traced to a private Chinese technology firm that does work for provincial clients, while OpenAI stopped short of saying the Chinese government ordered the campaigns.
- OpenAI judged both operations small and short-lived with little authentic engagement and found no evidence the content achieved meaningful breakout beyond the networks that posted it.
- Investigators found operators prompted ChatGPT in Simplified Chinese, used VPNs to access the tool from inside China, produced English- and Chinese-language outputs, and in some cases uploaded internal operational files describing tactics and evasion strategies into the model.
- The report has prompted calls from U.S. lawmakers and industry figures for probes and has sharpened debate over data-center projects as communities, regulators and companies weigh the growing risk that generative AI can be used to mass-produce influence content tied to real local concerns.