"This looks like a classic example of a foreign influence operation jumping onto the bandwagon of a genuine and pre-existing domestic debate and trying to manipulate it by using fake accounts posing as Americans,” said Ben Nimmo, principal investigator at OpenAI and author of the company's report.
OpenAI’s threat intelligence findings
OpenAI’s threat intelligence team tracked two distinct clusters of online activity that the company says “likely originated” in China. The firm described one cluster as an “anti-data center” operation and a second focused on tariffs and geopolitical influence. OpenAI did not directly attribute either campaign to the Chinese government or to actors working on its behalf, but said “many parts of the campaign and its tactics overlap with pre-established Chinese government propaganda campaigns online.”
OpenAI also cautioned that it is not a neutral party in the matter: the company has sought to raise “hundreds of millions of dollars” in funding to build data centers in the United States. The report, the company said, does not claim that anti-data center sentiment in the country is being driven or bolstered by foreign propaganda online.
“Data Center Bandwagon”: claims and provenance
One cluster, which OpenAI called “Data Center Bandwagon,” used ChatGPT to generate imagery and social media comments framing data center buildouts as responsible for raising electricity prices for Americans. OpenAI traced the anti-data center content to an unnamed Chinese technology company that holds multiple contracts with regional Chinese governments. The firm says the campaign posed as Americans on platforms including X and YouTube.
OpenAI judged this activity to have produced little traction outside its own amplification network. The campaigns received Bookings breakout scores of 1 and 2—ratings that OpenAI says indicate activity across one or more platforms but no evidence of meaningful engagement by targeted audiences.
Tariffs cluster and prompts that skewed messaging
The second cluster used ChatGPT to craft images and posts that characterized tariffs as a covert means for countries to exert control over the global technological landscape. According to the report, some originating prompts explicitly directed ChatGPT to include only U.S. President Donald Trump while excluding Chinese President Xi Jinping—despite Xi’s own use of tariffs. In addition to English-language posts, the tariff cluster produced short comments and comics in English, Italian, Japanese and traditional Chinese accusing the U.S. of prioritizing profits over loyalty to allies.
Tactics: ChatGPT, VPNs, operational security and platform evasion
OpenAI’s analysis identified a repeated operational pattern. Both clusters used VPNs to obscure origin, prompted ChatGPT in simplified Chinese, and requested outputs in both English and Chinese. Operators posed as Americans on major social platforms and, in at least one instance, fed internal documents into ChatGPT. OpenAI said the actors used ChatGPT to edit work reports that contained operational-security details about their social media campaigns, including goals of “establishing persistent and credible accounts, producing visually appealing content to expand audience reach in different regions and maintaining long term account viability by anticipating platform enforcement.”
Another report was run through ChatGPT that discussed how to leverage Facebook’s content ecosystem—groups, pages, hashtags, advertising tools, recommendation systems and reporting mechanisms—and strategies for evading Meta’s detection of coordinated inauthentic accounts. OpenAI also said the same network targeted the company on X with a false influence campaign alleging a widespread user data breach that Nimmo described as having “never happened.”
The company’s analysts noted that while AI tools can generate content at scale, many of the images and messages used by these actors “appear clunky or use overly direct messaging” and betray “a lack of familiarity with both the English language and internet virality.”
What this means for technologists, policymakers, and affected enterprises
- Technologists and security teams: Watch for operational-security metadata and edited internal reports surfaced via generative tools, and monitor for cross-language prompt patterns—simplified Chinese inputs requesting English and Chinese outputs—and VPN-enabled account activity posing as domestic users.
- Policymakers and regulators: Note that foreign influence actors may piggyback on existing domestic debates rather than create them; OpenAI’s use of Bookings breakout scores highlights the need for platform metrics that distinguish platform activity from meaningful audience engagement.
- Affected enterprises and procurement leaders: The anti-data center cluster was traced to an unnamed Chinese technology company with multiple regional government contracts; organizations planning infrastructure projects should be alert to coordinated messaging campaigns that could be amplified with AI-produced visuals and social posts.
OpenAI’s assessment is blunt: the debate about data centers already existed, and investigators “didn’t see any signs that [the influence operation] succeeded,” Nimmo said. But the report documents a clear pattern—use of ChatGPT to draft content, to edit operational documents, and to plan platform evasion—that raises a specific operational question for researchers and platforms: even when such campaigns fail to gain broad traction, how should platforms and defenders detect and block organized, AI-assisted attempts to impersonate domestic voices?
https://cyberscoop.com/openai-china-influence-campaign-chatgpt/
