OpenAI says fake accounts from China tried to turn Americans against data centers
OpenAI says China-linked accounts used ChatGPT to run covert influence campaigns on U.S. data centers and tech policy. The efforts reportedly got little authentic engagement.
Intelligence analysis by GPT-5.4 Mini
OpenAI says it found two China-linked clusters using ChatGPT to draft posts, images and fake-account tactics for covert influence campaigns aimed at U.S. data centers, tariffs and tech policy. The company says the efforts mostly failed to move public opinion.
OpenAI says some fake online accounts used its chatbot like a factory for making arguments, pictures, and pretend identities. It was like trying to sneak posters into a school debate without anyone knowing who made them.
Analysis
What OpenAI says it found
OpenAI says it identified two clusters of ChatGPT users it believes were likely based in China and involved in influence campaigns. The larger group, which it labeled the "Data Center Bandwagon," allegedly used ChatGPT to create English-language talking points, comic strips, and other material arguing that AI data centers increase electricity demand and raise consumer bills. OpenAI says the accounts posed as Americans from different backgrounds and posted the generated text and images on social media.
The company says it believes this group was part of a social media team at a private Chinese company serving local government clients. According to OpenAI, the users also uploaded a file describing their goals and strategies, including how to sway public opinion and create fake social accounts without being detected. OpenAI says the same group also targeted Chinese people living abroad and asked for insults aimed at Chinese dissidents and political commentators.
The second cluster
OpenAI says a second group generated comments and images criticizing U.S. tariffs and tech policy. That cluster focused on portraying the U.S. negatively, including claims that the country had betrayed allies. OpenAI says the users asked ChatGPT to write in several languages, including English, Italian, Japanese, and traditional Chinese, to reach Taiwanese audiences. They also requested that Xi Jinping not appear in the images they created.
OpenAI says the campaigns did not get much authentic engagement and do not appear to have meaningfully shifted public opinion. Even so, the company argues the effort is notable because it tried to quietly enter an active American debate about AI infrastructure while hiding who was behind it. The article also notes that concerns about electricity costs near AI data centers are real and already widely discussed, which may have given the campaign a topical opening.
Key points
- OpenAI says it found China-linked account clusters using ChatGPT for covert influence campaigns.
- One group focused on persuading Americans that AI data centers raise electricity demand and consumer bills.
- The accounts allegedly posed as Americans and posted AI-generated text and images on social media.
- A second cluster criticized U.S. tariffs and tech policy and targeted Taiwanese audiences in multiple languages.
- OpenAI says the campaigns drew little authentic engagement but mattered because they tried to enter an active policy debate covertly.
If OpenAI's detection holds up, it could make it harder for covert groups to use AI tools at scale without being noticed. Public debate about data centers and energy use could also become more informed if people separate real concerns from manipulation.
The same tools that help people write and create can also help influence campaigns move faster and look more local and believable. If the tactics improve, similar operations could keep targeting real policy debates while hiding their origin.
