OpenAI says it caught China using ChatGPT to turn you against tariffs and data centers — though it didn't do a great job

The efforts by these accounts, says OpenAI, show operators from China were testing narratives against AI infrastructure. The efforts, however, do not seem to have borne much fruit.

“The operation sought to exploit and amplify existing public concerns about energy prices and local impacts of data center development, but we found no evidence of meaningful breakout beyond its own activity,” wrote OpenAI.

The operation

OpenAI does not allow China to access its ChatGPT models, and the accounts that were banned used VPNs to access the platform. The company says the agents posed as Americans while using prompts in “simplified Chinese,” often requesting Chinese- and English-language outputs.

The cluster that focused on painting data centers as facilities that will increase electricity prices would have ChatGPT create comic strips, which the operators would post on social media using inauthentic accounts. The chatbot was also used to edit images and add text to generic images that supported their narrative.

OpenAI said it believes the group that gained access to its tool was a social media team from a Chinese company doing work on the government’s behalf. The group was hardly subtle, though, going so far as to upload a file to ChatGPT describing its objectives and strategies about how to sway public opinion and subtly establish fake social media accounts.

The group also reportedly used ChatGPT to harass Chinese dissidents.

Stirring a well-stirred pot

The campaigns focused on issues that were already front of mind for many people, which might be why they failed to find much traction with Americans.